betaFeedback

WorksMathematics

The JJ-Homomorphism From Localising The Cobordism Hypothesis

Abstract

Lurie's cobordism hypothesis classifies fully extended framed topological field theories valued in a symmetric monoidal (∞,n)(\infty,n)-category C\mathcal{C} with duals by the ∞\infty-groupoid Cfd\mathcal{C}^{\mathrm{fd}} of fully dualisable objects. This classification results in an O(n)O(n)-action on Cfd\mathcal{C}^{\mathrm{fd}}, where O(n)O(n) acts on the framed bordism category by change of framing. Lurie observed that for a connective spectrum, this action is ``a version of the classical JJ-homomorphism.'' We make this identification explicit: with respect to the right action of O(n)O(n) on framings, the O(n)O(n)-action on the localisation L(Bordnfr)≃SL(\mathrm{Bord}_n^{\mathrm{fr}}) \simeq S is precisely ⊖Jn\ominus J_n, the inverse of the unstable JJ-homomorphism Jn ⁣:O(n)→GL1(S)J_n \colon O(n) \to \mathrm{GL}_1(S). The identification of the action with the unstable JJ-homomorphism in the setting of smooth stacks is due to Grady–Pavlov; we give an independent proof in Lurie's original formulation of the cobordism hypothesis, using the abstract Pontryagin–Thom construction. The sign (i.e.\ whether the action is JnJ_n or ⊖Jn\ominus J_n) depends on the convention for the O(n)O(n)-action on framings, and is relevant for pin±\mathrm{pin}^\pm theories and hence for quantum systems with time-reversal symmetry. We compute the induced action on homotopy groups of the sphere spectrum in low degrees, yielding explicit invariants for invertible TFTs in dimensions 11, 22, and 33, and connect these to the Kitaev chain, the framing anomaly, the spin-statistics theorem, and anomaly inflow. We recover the Galatius–Madsen–Tillmann–Weiss theorem as a consequence of the cobordism hypothesis and relate the homotopy fixed-point and mapping-spectrum perspectives on invertible TFTs via the duality between tangential and normal framings.

Green Anguish

Figure 1. Green Anguish! Art by Sergio Pallante [24]

Introduction

The cobordism hypothesis, stated by Baez–Dolan [BD95] and sketched by Lurie [20], is a fundamental result in the theory of topological quantum field theories (TFTs). It asserts that a fully extended framed nn-dimensional TFT valued in a symmetric monoidal (∞,n)(\infty,n)-category CC with duals is completely determined by its value on the positively framed point. More precisely, the space TFTnfr(C)\mathrm{TFT}^{\mathrm{fr}}_{n}(C) of such theories is equivalent to the space1^{1} CfdC^{\mathrm{fd}} of fully dualisable objects of CC. Since O(n)O(n) acts on Bordnfr\mathrm{Bord}^{\mathrm{fr}}_{n} by change of framing this induces an O(n)O(n)-action on CfdC^{\mathrm{fd}}.

A central consequence of the cobordism hypothesis is the classification of invertible TFTs, which are those valued in a connective spectrum XX, such as the sphere spectrum SS or the maximal Picard subgroupoid of some symmetric monoidal (∞,n)(\infty,n)-category. Invertible TFTs describe the low-energy limits of symmetry-protected topological (SPT) phases and the anomalies of quantum field theories [11 Wit16]. The classification of such theories reduces, via the cobordism hypothesis, to understanding the O(n)O(n)-action on the localisation L(Bordnfr)L(\mathrm{Bord}^{\mathrm{fr}}_{n}) of the framed bordism category, which is a connective spectrum. The Galatius–Madsen–Tillmann–Weiss (GMTW) theorem [17] identifies L(BordnB)L(\mathrm{Bord}^{B}_{n}) with the Madsen–Tillmann spectrum MTBMTB for any tangential structure B→BO(n)B \to BO(n), providing a powerful computational tool.

Lurie observed that the O(n)O(n)-action on the space of fully dualisable objects of a Picard ∞\infty-groupoid is “a version of the classical JJ-homomorphism” from stable homotopy theory. The JJ-homomorphism Jn ⁣:O(n)→GL1(S)J_n\colon O(n) \to GL_1(S), defined by the one-point compactification of linear isometries, is one of the most studied maps in algebraic topology; its image in the stable homotopy groups of spheres is the subject of the Adams conjecture [4] and controls the chromatic height-1 layer of the sphere spectrum. This identification has also been established by Grady-Pavlov [16 Theorem 5.0.6] in the framework of geometric cobordism categories.

In this paper, we make Lurie’s observation completely explicit by giving a proof in Lurie’s original formulation of the cobordism hypothesis. Our main theorem identifies the O(n)O(n)-action on L(Bordnfr)≃SL(\mathrm{Bord}^{\mathrm{fr}}_{n}) \simeq S with the inverse of the unstable JJ-homomorphism, ⊖Jn\ominus J_n (Theorem 3.18), using the abstract Pontryagin–Thom collapse map. The sign depends on the convention for the O(n)O(n)-action on framings; with the right-action convention used in this paper, the action is ⊖Jn\ominus J_n. An equivalent result has been obtained by Grady-Pavlov [16 Theorem 5.0.6] in the framework of geometric cobordism categories; our proof works directly within Lurie’s formulation. We explain how the choice of convention (right vs. left action of O(n)O(n) on framings) determines whether the tangential action is ⊖Jn\ominus J_n or JnJ_n, and how this relates to the normal action appearing in the construction of Madsen–Tillmann spectra. This distinction, while invisible in low-degree homotopy groups, is relevant for the classification of pin±^{\pm} theories and the physics of time-reversal symmetry.

With the action identified, we compute its effect on the homotopy groups of SS in low degrees (Proposition 4.3), yielding explicit invariants for invertible TFTs in dimensions n=1,2,3n = 1, 2, 3. We then connect these computations to the physics of the Kitaev chain, the framing anomaly of chiral conformal field theories, the spin-statistics theorem, and anomaly inflow [28, 29, 11]. Finally, we recover the GMTW theorem as a formal consequence of the cobordism hypothesis, and discuss the relationship between the homotopy fixed point spectral sequence and the Adams spectral sequence for MTBMTB.

Our approach here offers two perspectives on the classification of invertible TFTs: the homotopy fixed point formulation XhBX^{hB} (where BB acts on XX via ⊖Jn\ominus J_n) and the mapping spectrum formulation Map⁡Sp(MTB,X)\operatorname{Map}_{\mathrm{Sp}}(MTB,X). The equivalence between these formulations is a manifestation of the duality between tangential and normal framings, and we explain how the ⊖\ominus sign in the GMTW formulation is absorbed by this duality. While the mapping spectrum perspective is well established in the literature [17, 26], the homotopy fixed point perspective—and its direct connection to the JJ-homomorphism—appears to be less widely known, and we develop it here.

Outline

This paper is organised as follows. § 2 recalls the background on bordism categories, the cobordism hypothesis, the JJ-homomorphism, and the localisation functor LL. § 3 constructs the O(n)O(n)-action on L(Bord⁡nfr)≃SL(\operatorname{Bord}^{\mathrm{fr}}_n) \simeq\mathbb{S} via the abstract Pontryagin–Thom collapse map, proves that it is given by ⊖Jn\ominus J_n (Theorem 3.18), and identifies L(Bord⁡nB)L(\operatorname{Bord}^{B}_n) with the Madsen–Tillmann spectrum MTBMTB (Proposition 3.22). § 4 computes the induced action on the low-degree homotopy groups of S\mathbb{S}. § 5 discusses the physical interpretations in terms of invertible field theories, anomalies, and symmetry-protected topological phases.

Notation and conventions

We introduce the notation to be used throughout the paper.

♭₁ Framings. A framing of a rank-nn vector bundle V→XV \to X is a trivialisation V≅RnV \cong\mathbb{R}^n, equivalently a bundle isomorphism Rn×X→V\mathbb{R}^n \times X \to V. A stable framing is a trivialisation of V⊕RN≅Rn+NV \oplus\mathbb{R}^N \cong\mathbb{R}^{n+N} for some N≥0N \ge0. For a kk-manifold WW with n≥kn \ge k, an nn-framing is a trivialisation TW⊕Rn−k≅RnTW \oplus\mathbb{R}^{n-k} \cong\mathbb{R}^n.

♭₂ Principal bundles and frame bundles. Principal GG-bundles have right actions. The frame bundle of a rank-nn vector bundle V→XV \to X is the principal GLn(R)GL_n(\mathbb{R})-bundle whose fibre over x∈Xx \in X is the collection of linear isomorphisms T ⁣:Rn→VxT \colon\mathbb{R}^n \to V_x, on which A∈GLn(R)A \in GL_n(\mathbb{R}) acts on the right by precomposition: T⋅A:=T∘AT \cdot A := T \circ A. A framing of VV is a section of the frame bundle, i.e. a map ϕ ⁣:Rn→V\phi\colon\mathbb{R}^n \to V (the inverse of a trivialisation V→RnV \to\mathbb{R}^n). The right action of g∈O(n)g \in O(n) on framings is by precomposition: g⋅ϕ:=ϕ∘gg \cdot\phi:= \phi\circ g.

♭₃ Madsen–Tillmann spectra. For a tangential structure ξ ⁣:B→BO(n)\xi\colon B \to BO(n), the localisation of the bordism category is L(Bord⁡nB)≃MTBL(\operatorname{Bord}^{B}_n) \simeq MTB, where MTBMTB is the Madsen–Tillmann spectrum (Definition 2.16).

♭₄ The identification O(n)op≅O(n)O(n)^{\mathrm{op}} \cong O(n). For any ∞\infty-group GG, inversion g↦g−1g \mapsto g^{-1} gives a canonical isomorphism G≅GopG \cong G^{\mathrm{op}}. If GG is abelian, the identity also gives an isomorphism G≅GopG \cong G^{\mathrm{op}}, and the two differ by the inversion automorphism. For O(n)O(n) (non-abelian for n≥2n \ge2), only the inversion isomorphism is available. For O=lim⁡nO(n)O = \lim_{n} O(n) (homotopy abelian), both are available, and they differ by an automorphism that we denote by ⊖\ominus. Throughout, we use the inversion isomorphism g↦g−1g \mapsto g^{-1} to identify O(n)op≅O(n)O(n)^{\mathrm{op}} \cong O(n) and Oop≅OO^{\mathrm{op}} \cong O, and this identification is compatible with the stabilisation maps O(n)→O(n+1)O(n) \to O(n+1) in the colimit. The commutativity identification is not used.

Acknowledgements

ASD and LS would like to thank Theo Johnson Freyd for reading a draft of the paper. ASD would like to thank Lennart Meier for encouraging him to think about problems arising in Physics, broadly speaking. ASD would also like to thank his PhD advisor Drew Heard and Shachar Carmeli for the support and encouragement. LS is grateful to Theo Johnson-Freyd and Lukas Müller for many discussions surrounding bordism (∞,n)(\infty,n)-categories and O(n)O(n)-actions.

LLM Declaration

The authors did not use any artificial intelligence tools in the preparation of this manuscript.

Setup

In this section, we recall the essential objects and theorems that form the main background for our work: cobordism categories and the cobordism hypothesis, the JJ-homomorphism and Thom spectra, and the localisation functor from symmetric monoidal (∞,n)(\infty,n)-categories to spectra. Throughout, we work with the theory of ∞\infty-categories, letting Cat⁡(∞,n)\operatorname{Cat}(\infty,n) denote the ∞\infty-category of (∞,n)(\infty,n)-categories, and Sp⁡\operatorname{Sp} the ∞\infty-category of spectra. We use Fun⁡⊗(−,−)\operatorname{Fun}^{\otimes}(-,-) to denote the space of symmetric monoidal functors.

Bordism categories and the cobordism hypothesis

Definition 2.1. Let ξ:B→BO(n)\xi:B\to BO(n) be a map of spaces, which we call a tangential structure. The bordism category Bord⁡nB\operatorname{Bord}_{n}^{B} is the symmetric monoidal (∞,n)(\infty,n)-category sketched as follows. For a complete definition, the reader is encouraged to consult [9] and [20 Section 2.2].

  1. Objects: compact 00-manifolds PP equipped with a BB-structure on TP⊕RnTP\oplus\mathbb{R}^{n}. Since TPTP is trivial (a 00-manifold is a finite set of points), a BB-structure is a map from each point to the homotopy fibre of ξ\xi, i.e., a lift of the constant map to Rn\mathbb{R}^{n} through ξ\xi.

  1. 1-morphisms: a 11-morphism from P0P_{0} to P1P_{1} is a compact 11-manifold WW with a BB-structure on TW⊕Rn−1TW\oplus\mathbb{R}^{n-1} and an identification ∂W=P0⊔P1\partial W=P_{0}\sqcup P_{1} compatible with the BB-structures using in- and outgoing normal vectors on P0P_{0} and P1P_{1} respectively.

  1. kk-morphisms, 2≤k≤n2\leq k\leq n: a kk-morphism between two (k−1)(k-1)-morphisms is a compact kk-manifold WW with corners, a BB-structure on TW⊕Rn−kTW\oplus\mathbb{R}^{n-k} and an identification of its boundary with the incoming and outgoing (k−1)(k-1)-morphisms.

  1. (n+1)(n+1)-morphisms: diffeomorphisms of nn-manifolds preserving the BB-structure.

  1. Higher morphisms: Isotopies of BB-diffeomorphisms.

  1. Composition: glueing of bordisms along boundaries.

The symmetric monoidal structure is given by disjoint union.

Definition 2.2. For n≥kn\geq k, an nn-framing on a kk-manifold WW (potentially with corners) is a BB-structure on TW⊕Rn−kTW\oplus\mathbb{R}^{n-k} for B=ptB=\mathrm{pt}.

Recall that the frame bundle of an nn-dimensional vector bundle V→XV\to X is the principal GLn(R)GL_{n}(\mathbb{R})-bundle whose fibre over x∈Xx\in X is the collection of linear isomorphisms T:Rn→VxT:\mathbb{R}^{n}\to V_{x} on which A∈GLn(R)A\in GL_{n}(\mathbb{R}) acts by right composition. A section of the frame bundle is a basis of all the VxV_{x} continuous in xx. An nn-framing on WW is a trivialisation of TW⊕Rn−kTW\oplus\mathbb{R}^{n-k}, which therefore corresponds to a trivialisation of its frame bundle.

Remark 2.3. The data of an nn-framing on a kk-manifold for n>kn>k is typically weaker than a kk-framing. For example, SkS^{k} always has a (k+1)(k+1)-framing but only a kk-framing for k=0,1,3k=0,1,3 and 77.

Definition 2.4. When B=ptB=\mathrm{pt} with the unique map pt→BO(n)\mathrm{pt}\to BO(n), we obtain the fully framed bordism category Bord⁡nfr\operatorname{Bord}_{n}^{\mathrm{fr}}.

Remark 2.5. The GLn(R)GL_{n}(\mathbb{R})-action on the frame bundle yields a continuous right action of O(n)O(n) on Bord⁡nfr\operatorname{Bord}_{n}^{\mathrm{fr}} by change of framing. More precisely, there is a map of E1\mathbb{E}_{1}-spaces

αCH:O(n)op→Aut⁡Cat⁡(∞,n)E∞(Bord⁡nfr).\alpha_{\mathrm{CH}}:O(n)^{\mathrm{op}}\to\operatorname{Aut}_{\operatorname{Cat}_{(\infty,n)}^{\mathbb{E}_{\infty}}}\left(\operatorname{Bord}_{n}^{\mathrm{fr}}\right).

Alternatively, one can describe the delooping as a map

BO(n)→BAut⁡(Bord⁡nfr)↪Cat⁡(∞,n)E∞,BO(n)\to B\operatorname{Aut}\left(\operatorname{Bord}_{n}^{\mathrm{fr}}\right)\hookrightarrow\operatorname{Cat}_{(\infty,n)}^{\mathbb{E}_{\infty}},

which is the input for the cobordism hypothesis.

A fully extended framed nn-dimensional topological field theory valued in a symmetric monoidal (∞,n)(\infty,n)-category CC is a symmetric monoidal functor Z:Bord⁡nfr→CZ:\operatorname{Bord}^{\mathrm{fr}}_n\to C. We denote the space of such functors by TFT⁡nfr(C):=Fun⁡⊗(Bord⁡nfr,C)\operatorname{TFT}^{\mathrm{fr}}_n(C):=\operatorname{Fun}^{\otimes}(\operatorname{Bord}^{\mathrm{fr}}_n,C). The foundational result in this subject is the cobordism hypothesis, stated by Baez-Dolan [BD95].

Theorem 2.6 (Cobordism Hypothesis, [20, 16]). Let CC be a symmetric monoidal (∞,n)(\infty,n)-category with duals. Evaluation at the positively framed point pt+∈Bord⁡nfr\mathrm{pt}^{+}\in\operatorname{Bord}^{\mathrm{fr}}_n induces an equivalence of spaces

TFT⁡nfr(C)≃Cfd\operatorname{TFT}^{\mathrm{fr}}_n(C)\simeq C^{\mathrm{fd}}

where CfdC^{\mathrm{fd}} denotes the underlying ∞\infty-groupoid of fully dualisable objects in CC.

Remark 2.7. The above theorem gives an O(n)O(n)-action on CfdC^{\mathrm{fd}} from the O(n)O(n)-action on Bord⁡nfr\operatorname{Bord}^{\mathrm{fr}}_n via the equivalence. The ∞\infty-category Fun⁡(BO(n),S)\operatorname{Fun}(BO(n),\mathcal{S}) of spaces with an O(n)O(n)-action remembers all higher coherence of the O(n)O(n)-equivariance. Concretely, the equivalence TFT⁡nfr(C)≃Cfd\operatorname{TFT}^{\mathrm{fr}}_n(C)\simeq C^{\mathrm{fd}} intertwines the two O(n)O(n)-actions via a specified homotopy, together with a compatible family of higher homotopies witnessing the O(n)O(n)-coherence. At the level of points, this implies that for any g∈O(n)g\in O(n), the diagram

TFT⁡nfr(C)→≃Cfdg⋅↓↓g⋅TFT⁡nfr(C)→≃Cfd\begin{CD} \operatorname{TFT}^{\mathrm{fr}}_n(C) @>\simeq>> C^{\mathrm{fd}} \\ @V{g\cdot}VV @VV{g\cdot}V \\ \operatorname{TFT}^{\mathrm{fr}}_n(C) @>\simeq>> C^{\mathrm{fd}} \end{CD}

commutes, where the left vertical map precomposes with the change-of-framing functor Bord⁡nfr→gBord⁡nfr\operatorname{Bord}^{\mathrm{fr}}_n\xrightarrow{g}\operatorname{Bord}^{\mathrm{fr}}_n, and the right vertical map is the action constructed by Lurie [20].

Using the equivalence Bord⁡nB≃colim⁡BBord⁡nfr\operatorname{Bord}^{B}_n\simeq\operatorname*{colim}_{B}\operatorname{Bord}^{\mathrm{fr}}_n (the colimit of the BB-action via ξ\xi) and the fact that Fun⁡⊗(−,C)\operatorname{Fun}^{\otimes}(-,C) converts colimits to limits, one gets the corollary:

Corollary 2.8. For any CC as above and any map ξ:B→BO(n)\xi:B\to BO(n),

TFT⁡nB(C)≃(Cfd)hB,\operatorname{TFT}^{B}_n(C)\simeq\left(C^{\mathrm{fd}}\right)^{hB},

where (−)hB(-)^{hB} denotes homotopy fixed points for the action of BB on CfdC^{\mathrm{fd}} through the composition B→ξBO(n)→Aut⁡(Cfd)B\xrightarrow{\xi}BO(n)\to\operatorname{Aut}(C^{\mathrm{fd}}).

Units, the J-homomorphism, and Thom spectra

We denote by Sp⁡\operatorname{Sp} the symmetric monoidal ∞\infty-category of spectra. Its monoidal unit is the sphere spectrum S\mathbb{S}. By a classical theorem of May [22 Theorem 14.4, Corollary 14.5], and its ∞\infty-categorical refinement [21 Remark 5.2.6.26], there is an equivalence between the ∞\infty-category of connective spectra Sp⁡≥0\operatorname{Sp}_{\geq0} and the ∞\infty-category of grouplike E∞E_{\infty}-spaces. For a spectrum XX, we let Ω∞X∈S\Omega^{\infty}X\in\mathcal{S} denote its underlying (infinite loop) space. Recall that an E∞E_{\infty}-ring spectrum is a commutative algebra object in Sp⁡\operatorname{Sp}.

Definition 2.9. Let RR be an E∞E_{\infty}-ring spectrum.

♮1 The higher group of units gl⁡1(R)\operatorname{gl}_{1}(R) is defined as the E∞E_{\infty}-group of automorphisms of the unit R∈Mod⁡RR\in\operatorname{Mod}_{R} in the category of right RR-modules.

♮2 The Picard groupoid Pic⁡(R)\operatorname{Pic}(R) of RR is the E∞E_{\infty}-group of invertible modules and invertible module maps between them.

We will typically identify gl⁡1(R)\operatorname{gl}_{1}(R) and Pic⁡(R)\operatorname{Pic}(R) with their corresponding connective spectra. We have ΩPic⁡(R)=gl⁡1(R)\Omega\operatorname{Pic}(R)=\operatorname{gl}_{1}(R). If GL⁡1(R):=Ω∞gl⁡1(R)\operatorname{GL}_{1}(R):=\Omega^{\infty}\operatorname{gl}_{1}(R) is the space of units [3 Definition 2.1], there is pullback

\begin{CD} \mathrm{GL}_1(R) @>>> \Omega^\infty R \\ @VVV @VV{\pi_0}V \\ (\pi_0R)^\times@>>> \pi_0R \end{CD}

in the ∞\infty-category S\mathcal{S} [1 Proposition 6.2].

Since SS is the monoidal unit in Sp\mathrm{Sp}, it has a canonical E∞E_\infty-structure and gl1(S)=Aut⁡Sp(S)\mathrm{gl}_1(S)=\operatorname{Aut}_{\mathrm{Sp}}(S). For the sphere spectrum SS, we have π0(S)≅Z\pi_0(S)\cong\mathbb{Z} and therefore (π0S)×≅{±1}≅Z/2(\pi_0S)^\times\cong\{\pm1\}\cong\mathbb{Z}/2. The space GL1(S)\mathrm{GL}_1(S) has two connected components, and its homotopy groups are

πk(GL1(S))≅{Z/2for k=0,πk(S)for k≥1.\pi_k(\mathrm{GL}_1(S))\cong \begin{cases} \mathbb{Z}/2 & \text{for } k=0,\\ \pi_k(S) & \text{for } k\geq1. \end{cases}

Definition 2.10. For a fixed dimension nn, the unstable JJ-homomorphism is the map of E1E_1-spaces

Jn ⁣:O(n)⟶GL1(S)J_n\colon O(n)\longrightarrow\mathrm{GL}_1(S)

defined as follows. A point g∈O(n)g\in O(n) defines a linear isometry Rn→Rn\mathbb{R}^n\to\mathbb{R}^n, whose one-point compactification is a based map Sn→SnS^n\to S^n, an element of the nn-fold loop space ΩnSn\Omega^nS^n. We set Jn(g)J_n(g) to be the image under

O(n)↪ΩnSn→stabΩ∞S,O(n)\hookrightarrow\Omega^nS^n\xrightarrow{\mathrm{stab}}\Omega^\infty S,

where the first map is the classical unstable JJ-homomorphism [6 pp. 291–293] and stab\mathrm{stab} is the stabilisation map ΩnSn→lim⁡kΩkSk=:Ω∞S\Omega^nS^n\to\lim\limits_{k}\Omega^kS^k=:\Omega^\infty S. The image of the composition lies in the units GL1(S)⊂Ω∞S\mathrm{GL}_1(S)\subset\Omega^\infty S by inspection of π0\pi_0: an element g∈O(n)g\in O(n) maps to a based map Sn→SnS^n\to S^n of degree det⁡(g)=±1\det(g)=\pm1, whose stabilisation has degree ±1\pm1 in π0(S)≅Z\pi_0(S)\cong\mathbb{Z}, hence lies in the units π0(S)×={±1}⊂GL1(S)\pi_0(S)^\times=\{\pm1\}\subset\mathrm{GL}_1(S). Note that the stabilisation map ΩnSn→Ω∞S\Omega^nS^n\to\Omega^\infty S is not a map to GL1(S)\mathrm{GL}_1(S) on all of ΩnSn\Omega^nS^n: the components of degree ≠±1\ne\pm1 do not land in the units.

The unstable JJ-homomorphisms JnJ_n are compatible with the stabilisation inclusions O(n)↪O(n+1)O(n)\hookrightarrow O(n+1), in the sense that the diagram

O(n)→JnGL1(S)↓↓O(n+1)→Jn+1GL1(S)\begin{CD} O(n) @>{J_n}>> \mathrm{GL}_1(S) \\ @VVV @VVV \\ O(n+1) @>{J_{n+1}}>> \mathrm{GL}_1(S) \end{CD}

commutes up to homotopy. Passing to the colimit, we obtain the stable JJ-homomorphism

J ⁣:O=lim⁡nO(n)⟶GL1(S).J\colon O=\lim\limits_n O(n)\longrightarrow\mathrm{GL}_1(S).

Equivalently, at the level of classifying spaces, this gives BJ ⁣:BO→BGL1(S)BJ\colon BO\to B\mathrm{GL}_1(S), which classifies stable spherical fibrations.

For the purpose of defining Thom spectra, it is convenient to give a more conceptual construction of JJ that packages the whole family {Jn}n\{J_n\}_n into a single map. Let Vec≃⨆n≥0BO(n)\mathrm{Vec}\simeq\bigsqcup_{n\geq0}BO(n) denote the E∞E_\infty-groupoid of real vector spaces and orthogonal maps, under Whitney sum ⊕\oplus. One-point compactification with basepoint at infinity defines a symmetric monoidal functor

Vec⟶S∗,V↦SV.(1)\mathrm{Vec}\longrightarrow\mathcal{S}_*,\qquad V\mapsto S^V. \tag*{(1)}

where we give pointed spaces the Cartesian symmetric monoidal structure. Composing (1) with the symmetric monoidal left adjoint Σ∞:S∗→Sp⁡\Sigma^\infty: \mathcal{S}_* \to\operatorname{Sp} of Ω∞\Omega^\infty, we obtain a symmetric monoidal functor (Vec⁡,⊕)→(Sp⁡,∧)(\operatorname{Vec}, \oplus) \to(\operatorname{Sp}, \wedge).

Since SV∧S−V≃SS^V \wedge S^{-V} \simeq S and O(V)O(V) maps to automorphisms of SVS^V, this functor lands in the non-full subcategory Pic⁡(S)↪Sp⁡\operatorname{Pic}(S) \hookrightarrow\operatorname{Sp} of invertible SS-modules. Since Pic⁡(S)\operatorname{Pic}(S) is grouplike, the universal property of the group completion ko=BO×Zko = BO \times\mathbb{Z} of Vec⁡\operatorname{Vec} implies that the composition factors through a map

J:BO×Z⟶Pic⁡(S).J : BO \times\mathbb{Z} \longrightarrow\operatorname{Pic}(S).

The restriction of JJ to the component BO(n)⊂BO×ZBO(n) \subset BO \times\mathbb{Z} recovers the unstable JnJ_n: on the level of π0\pi_0, both send a rank-nn vector space to ΣnS\Sigma^n S, and the induced action of O(n)O(n) on ΣnS\Sigma^n S is the same. Thus the conceptual construction above is a repackaging of the family {Jn}n\{J_n\}_n into a single stable map.

Definition 2.11. The stable JJ-homomorphism is the map

J:BO×Z⟶Pic⁡(S)J : BO \times\mathbb{Z} \longrightarrow\operatorname{Pic}(S)

constructed above.

Definition 2.12. A stable vector bundle over XX is a map X→BO×ZX \to BO \times\mathbb{Z}. If X→BO(n)X \to BO(n) is a map classifying a rank-nn vector bundle, its stabilisation is its composition with the group completion map BO(n)→BO×ZBO(n) \to BO \times\mathbb{Z} that includes into the {n}⊆Z\{n\} \subseteq\mathbb{Z} component.

Remark 2.13. Observe that in our definition, a rank-nn vector bundle includes into the {n}⊆Z\{n\} \subseteq\mathbb{Z} component, and stable vector bundles can have a nonzero virtual rank. Because JJ is defined using the universal property of the group completion, we have J(V−W)=SV∧S−WJ(V-W) = S^V \wedge S^{-W} for vector spaces V,WV,W.

Definition 2.14. The Thom spectrum XVX^V of a stable vector bundle X→BO×ZX \to BO \times\mathbb{Z} is the colimit of

X⟶BO×Z→JPic⁡(S)⟶Sp⁡.X \longrightarrow BO \times\mathbb{Z} \xrightarrow{J} \operatorname{Pic}(S) \longrightarrow\operatorname{Sp}.

Remark 2.15. A functor X→S∗X \to\mathcal{S}_* corresponds to a fibration with total space EE, its colimit. For X→Vec⁡X \to\operatorname{Vec} a vector bundle, the Thom space can be computed as

colim⁡(X→S∗→S)⨿Xpt⁡=cofib⁡(s:X→E),\operatorname{colim}(X \to\mathcal{S}_* \to\mathcal{S}) \amalg_X \operatorname{pt} = \operatorname{cofib}(s:X \to E),

where E=colim⁡(X→S∗→S)E = \operatorname{colim}(X \to\mathcal{S}_* \to\mathcal{S}) is the total space of the associated spherical fibration, and ss its canonical section. This connects Definition 2.14 with the usual definition of Thom space.

Definition 2.16. For a tangential structure ξ:B→BO(n)\xi: B \to BO(n), the Madsen–Tillmann spectrum MTBMTB is the Thom spectrum of the virtual normal bundle, i.e. the colimit of

B→⊖(ι,n)∘ξBO×Z→JPic⁡(S)⟶Sp⁡,B \xrightarrow{\ominus(\iota,n)\circ\xi} BO \times\mathbb{Z} \xrightarrow{J} \operatorname{Pic}(S) \longrightarrow\operatorname{Sp},

where ι:BO(n)→BO\iota: BO(n) \to BO is the stabilisation map, n:BO(n)→Zn: BO(n) \to\mathbb{Z} is the constant map at nn, and ⊖\ominus is the inverse in the grouplike E∞E_\infty-structure on BO×ZBO \times\mathbb{Z} coming from Whitney sum. Equivalently, under the identification of Definition 2.14, MTB=B−ξMTB = B^{-\xi}, the Thom spectrum of the virtual bundle −ξ-\xi of rank −n-n. A closed nn-manifold MM with a BB-structure gives a map S→MTBS \to MTB via the Pontryagin–Thom construction, so that MTBMTB is the spectrum of BB-structured bordism classes in degree 00.

Example 2.17. Note that BO(n)→Pic⁡(S)BO(n) \to\operatorname{Pic}(S) maps to ΣnS\Sigma^n S. Therefore, the Thom spectrum of an nn-dimensional trivial bundle over XX is the colimit over the trivial diagram indexed by XX, which is ΣnΣ+∞X\Sigma^n \Sigma^\infty_+ X.

The localisation functor LL

The ∞\infty-category of (∞,n)(\infty,n)-categories and the localisation functor

We recall the recursive definition of the ∞\infty-category of (∞,n)(\infty,n)-categories and the adjunctions between the various levels. The primary reference is [14].

Definition 2.18. We define Cat⁡(∞,n)\operatorname{Cat}_{(\infty,n)} recursively by

Cat⁡(∞,0):=S,Cat⁡(∞,n):=Cat⁡[Cat⁡(∞,n−1)].\operatorname{Cat}_{(\infty,0)} := \mathcal{S}, \qquad\operatorname{Cat}_{(\infty,n)} := \operatorname{Cat}[\operatorname{Cat}_{(\infty,n-1)}].

where S\mathcal{S} is the ∞\infty-category of spaces and Cat⁡[−]\operatorname{Cat}[-] is the construction of [14 §5] taking a monoidal ∞\infty-category V\mathcal{V} to the ∞\infty-category Cat⁡[V]\operatorname{Cat}[\mathcal{V}] of small V\mathcal{V}-enriched ∞\infty-categories.

By [14 Theorem 5.4.6], we have Cat⁡(∞,1)≃Cat⁡∞\operatorname{Cat}_{(\infty,1)} \simeq\operatorname{Cat}_{\infty}, and the construction is Cartesian presentably symmetric monoidal at every level [14 Proposition 5.7.16].

There are two natural adjunctions between consecutive levels:

♮1\natural_1 Core / inclusion. The core functor

(−)core⁡n−1:Cat⁡(∞,n)→Cat⁡(∞,n−1)(-)^{\operatorname{core}_{n-1}} : \operatorname{Cat}_{(\infty,n)} \to\operatorname{Cat}_{(\infty,n-1)}

is right adjoint to the fully faithful inclusion

ιn:Cat⁡(∞,n−1)↪Cat⁡(∞,n)\iota_n : \operatorname{Cat}_{(\infty,n-1)} \hookrightarrow\operatorname{Cat}_{(\infty,n)}

of (∞,n−1)(\infty,n-1)-categories as (∞,n)(\infty,n)-categories with only invertible nn-morphisms.

♮2\natural_2 Localisation / inclusion. The groupoid-completion functor

Ln−1:Cat⁡(∞,n)→Cat⁡(∞,n−1)L_{n-1} : \operatorname{Cat}_{(\infty,n)} \to\operatorname{Cat}_{(\infty,n-1)}

is left adjoint to ιn\iota_n. It universally inverts all nn-morphisms.

The composite L0:Cat⁡(∞,n)→SL_0 : \operatorname{Cat}_{(\infty,n)} \to\mathcal{S} is the ∞\infty-groupoid completion functor.

Lemma 2.19. The construction Cat⁡[−]\operatorname{Cat}[-] preserves adjunctions: if F:C⇄D:GF : \mathcal{C} \rightleftarrows\mathcal{D} : G is an adjunction of monoidal ∞\infty-categories, then

Cat⁡[F]:Cat⁡[C]⇄Cat⁡[D]:Cat⁡[G]\operatorname{Cat}[F] : \operatorname{Cat}[\mathcal{C}] \rightleftarrows\operatorname{Cat}[\mathcal{D}] : \operatorname{Cat}[G]

is an adjunction of ∞\infty-categories.

Proof. By [14 Proposition 5.7.8 and Proposition 5.7.16], the construction Cat⁡[−]\operatorname{Cat}[-] assembles into a functor

Cat⁡[−]:CAlg⁡(Pr⁡L)→CAlg⁡(Pr⁡L).\operatorname{Cat}[-] : \operatorname{CAlg}(\operatorname{Pr}^{\mathrm{L}}) \to\operatorname{CAlg}(\operatorname{Pr}^{\mathrm{L}}).

Since Pr⁡L\operatorname{Pr}^{\mathrm{L}} is the ∞\infty-category of presentable ∞\infty-categories and left adjoints, such a functor sends left adjoints to left adjoints and right adjoints to right adjoints. Therefore, it preserves adjunctions. □\square

Lemma 2.20. The construction Cat⁡[−]\operatorname{Cat}[-] preserves finite products: if C,D\mathcal{C}, \mathcal{D} are monoidal ∞\infty-categories, then

Cat⁡[C×D]≃Cat⁡[C]×Cat⁡[D].\operatorname{Cat}[\mathcal{C} \times\mathcal{D}] \simeq\operatorname{Cat}[\mathcal{C}] \times\operatorname{Cat}[\mathcal{D}].

More generally, if C→D\mathcal{C} \to\mathcal{D} is a symmetric monoidal functor, then so is Cat⁡[C]→Cat⁡[D]\operatorname{Cat}[\mathcal{C}] \to\operatorname{Cat}[\mathcal{D}].

Proof. By [14 Proposition 5.7.16], the construction Cat⁡[−]\operatorname{Cat}[-] is symmetric monoidal with respect to the Cartesian symmetric monoidal structure on Cat⁡(∞,n−1)\operatorname{Cat}_{(\infty,n-1)}. Since the monoidal structure is Cartesian, preserving it means preserving products. The second statement follows from the same citation. □\square

Remark 2.21. Combining Lemmas 2.19 and 2.20, we obtain by induction that:

♮1\natural_1 The adjunctions Ln−1⊣ιnL_{n-1} \dashv\iota_n are symmetric monoidal adjunctions, for every n≥1n \ge1.

♮2\natural_2 The ∞\infty-groupoid completion functor L0:Cat⁡(∞,n)→SL_0 : \operatorname{Cat}_{(\infty,n)} \to\mathcal{S} is symmetric monoidal.

This is the key inductive input for the symmetric monoidality of the localisation functor LL (Lemma 2.23).

Let Cat⁡(∞,n)E∞\operatorname{Cat}^{E_\infty}_{(\infty,n)} denote the ∞\infty-category of symmetric monoidal (∞,n)(\infty,n)-categories and symmetric monoidal functors between them. There is a natural fully faithful functor

ι:Sp⁡≥0↪Cat⁡(∞,n)E∞\iota:\operatorname{Sp}_{\geq0}\hookrightarrow\operatorname{Cat}^{E_\infty}_{(\infty,n)}

defined as follows. A connective spectrum XX determines a grouplike E∞E_\infty-space Ω∞X\Omega^\infty X, which is in particular a symmetric monoidal (∞,0)(\infty,0)-category (an ∞\infty-groupoid). We view this as a symmetric monoidal (∞,n)(\infty,n)-category with only invertible kk-morphisms for all 1≤k≤n1\leq k\leq n. Under the homotopy hypothesis, this is the ∞\infty-category obtained from Ω∞X\Omega^\infty X by taking the fundamental ∞\infty-groupoid in which paths as 11-morphisms, homotopies as 22-morphisms, and so on. All objects of ι(X)\iota(X) are fully dualisable because the monoidal structure is grouplike: every object xx has an inverse x−1x^{-1} under the tensor product, which serves as its dual.

The ∞\infty-category Cat⁡(∞,n)E∞\operatorname{Cat}^{E_\infty}_{(\infty,n)} is presentable [27 Proposition 5.1.7 + Definition 5.3.6], and ι\iota is an accessible right adjoint. Therefore, by the adjoint functor theorem for presentable ∞\infty-categories [23 Theorem 4.1.1], ι\iota admits a left adjoint

L:Cat⁡(∞,n)E∞⟶Sp⁡≥0.L:\operatorname{Cat}^{E_\infty}_{(\infty,n)}\longrightarrow\operatorname{Sp}_{\geq0}.

We call LL the localisation functor. On a symmetric monoidal (∞,n)(\infty,n)-category C\mathcal{C}, the functor LL universally forces all kk-morphisms for 0≤k≤n0\leq k\leq n to become equivalences. Since k=0k=0 is included, this also performs group completion with respect to the symmetric monoidal structure. Concretely, the underlying infinite loop space of L(C)L(\mathcal{C}) is the group completion of the localisation L0CL_0\mathcal{C} of C\mathcal{C} with respect to kk-morphisms for 0<k≤n0<k\leq n. We write

Lk:Cat⁡(∞,n)E∞⟶Cat⁡(∞,k)E∞L_k:\operatorname{Cat}^{E_\infty}_{(\infty,n)}\longrightarrow\operatorname{Cat}^{E_\infty}_{(\infty,k)}

for the level-wise localisation functor (for k<nk<n), i.e. the functor that inverts all jj-morphisms for j>kj>k, and

L:Cat⁡(∞,n)E∞⟶Sp⁡≥0L:\operatorname{Cat}^{E_\infty}_{(\infty,n)}\longrightarrow\operatorname{Sp}_{\geq0}

for the global localisation functor, left adjoint to ι\iota as above. The global LL factors as the composite

L:Cat⁡(∞,n)E∞→Ln−1Cat⁡(∞,n−1)E∞→Ln−2⋯→L0Cat⁡(∞,0)E∞=SE∞→group completionSp⁡≥0,L:\operatorname{Cat}^{E_\infty}_{(\infty,n)} \xrightarrow{L_{n-1}}\operatorname{Cat}^{E_\infty}_{(\infty,n-1)} \xrightarrow{L_{n-2}}\cdots \xrightarrow{L_0}\operatorname{Cat}^{E_\infty}_{(\infty,0)} =S^{E_\infty} \xrightarrow{\text{group completion}}\operatorname{Sp}_{\geq0},

where the intermediate localisations are the level-wise ones restricted to the symmetric monoidal subcategory.

Lemma 2.22. Let k<nk<n. The level-wise localisation Lk:Cat⁡(∞,n)→Cat⁡(∞,k)L_k:\operatorname{Cat}_{(\infty,n)}\to\operatorname{Cat}_{(\infty,k)} preserves finite products.

Proof. The ∞\infty-category Cat⁡(∞,n)\operatorname{Cat}_{(\infty,n)} has exponential objects given by functor (∞,n)(\infty,n)-categories: if C∈Cat⁡(∞,n)\mathcal{C}\in\operatorname{Cat}_{(\infty,n)} and D∈Cat⁡(∞,k)\mathcal{D}\in\operatorname{Cat}_{(\infty,k)}, then CD∈Cat⁡(∞,k)\mathcal{C}^{\mathcal{D}}\in\operatorname{Cat}_{(\infty,k)} (the exponentials are computed levelwise, and LkL_k only affects the top levels). It follows for C1,C2∈Cat⁡(∞,n)\mathcal{C}_1,\mathcal{C}_2\in\operatorname{Cat}_{(\infty,n)} and D∈Cat⁡(∞,k)\mathcal{D}\in\operatorname{Cat}_{(\infty,k)} that

Hom⁡Cat⁡(∞,k)(Lk(C1×C2),D)≃Hom⁡Cat⁡(∞,n)(LkC1,DC2)≃Hom⁡Cat⁡(∞,n)(C1,DC2)≃Hom⁡Cat⁡(∞,n)(C1×C2,D).\operatorname{Hom}_{\operatorname{Cat}_{(\infty,k)}}(L_k(\mathcal{C}_1\times\mathcal{C}_2),\mathcal{D})\simeq\operatorname{Hom}_{\operatorname{Cat}_{(\infty,n)}}(L_k\mathcal{C}_1,\mathcal{D}^{\mathcal{C}_2})\simeq\operatorname{Hom}_{\operatorname{Cat}_{(\infty,n)}}(\mathcal{C}_1,\mathcal{D}^{\mathcal{C}_2})\simeq\operatorname{Hom}_{\operatorname{Cat}_{(\infty,n)}}(\mathcal{C}_1\times\mathcal{C}_2,\mathcal{D}).

By Yoneda, the unit map C1×C2→LkC1×C2\mathcal{C}_1\times\mathcal{C}_2\to L_k\mathcal{C}_1\times\mathcal{C}_2 becomes an equivalence after applying LkL_k. Repeating the argument for the other factor proves the same for C1×C2→LkC1×LkC2\mathcal{C}_1\times\mathcal{C}_2\to L_k\mathcal{C}_1\times L_k\mathcal{C}_2. Since LkC1×LkC2∈Cat⁡(∞,k)L_k\mathcal{C}_1\times L_k\mathcal{C}_2\in\operatorname{Cat}_{(\infty,k)}, we obtain

Lk(C1×C2)≃Lk(LkC1×LkC2)≃LkC1×LkC2,L_k(\mathcal{C}_1\times\mathcal{C}_2)\simeq L_k(L_k\mathcal{C}_1\times L_k\mathcal{C}_2)\simeq L_k\mathcal{C}_1\times L_k\mathcal{C}_2,

so LkL_k preserves binary products. The terminal object is preserved since Lk(pt)≃ptL_k(\mathrm{pt})\simeq\mathrm{pt} (the terminal (∞,n)(\infty,n)-category has only one object and only identity morphisms, hence is already local). Therefore LkL_k preserves finite products. □

Lemma 2.23. The global localisation functor L ⁣:Cat⁡(∞,n)E∞→Sp⁡≥0L\colon\operatorname{Cat}^{\mathrm{E}_{\infty}}_{(\infty,n)} \to\operatorname{Sp}_{\geq0} is symmetric monoidal, with respect to the Cartesian symmetric monoidal structure on Cat⁡(∞,n)E∞\operatorname{Cat}^{\mathrm{E}_{\infty}}_{(\infty,n)} and the smash product on Sp⁡≥0\operatorname{Sp}_{\geq0}.

Proof. The functor ι ⁣:Sp⁡≥0↪Cat⁡(∞,n)E∞\iota\colon\operatorname{Sp}_{\geq0} \hookrightarrow\operatorname{Cat}^{\mathrm{E}_{\infty}}_{(\infty,n)} factors as

Sp⁡≥0→ι1S→ι2Cat⁡(∞,n)E∞.\operatorname{Sp}_{\geq0} \xrightarrow{\iota_1} \mathcal{S} \xrightarrow{\iota_2} \operatorname{Cat}^{\mathrm{E}_{\infty}}_{(\infty,n)}.

where ι1\iota_1 includes connective spectra as grouplike E∞\mathrm{E}_{\infty}-spaces and ι2\iota_2 includes spaces as (∞,n)(\infty,n)-categories with only invertible morphisms. The left adjoint to ι2\iota_2 is the ∞\infty-groupoid completion functor, which is symmetric monoidal by Remark 2.21. The left adjoint to ι1\iota_1 is group completion, which is symmetric monoidal (group completion is a colimit of symmetric monoidal functors, so it inherits a symmetric monoidal structure). Therefore, LL, being the composite of these two symmetric monoidal left adjoints, is symmetric monoidal. □

Remark 2.24. If CC is a symmetric monoidal (∞,n)(\infty,n)-category with duals, then for every k>0k > 0, any kk-morphism has a left and a right adjoint [20 Definition 2.3.13+Definition 2.3.16] and is therefore already invertible in L1(C)L_1(C). Consequently, L1(C)L_1(C) is already the full localisation L(C)L(C). This is precisely the situation for the bordism category Bord⁡nB\operatorname{Bord}^{B}_{n} when n>0n > 0.

Localisation of the Cobordism Hypothesis Action

Recovering the sphere spectrum from abstract nonsense

One way to phrase the cobordism hypothesis is to say that Bord⁡nfr\operatorname{Bord}^{\mathrm{fr}}_{n} is the free symmetric monoidal (∞,n)(\infty,n)-category with duals on a single object. Geometrically, this object is the connected 00-dimensional manifold with its positive framing. Since the localisation functor LL is a left adjoint, it preserves colimits, and therefore preserves the property of being free. Therefore, we expect L(Bord⁡nfr)L(\operatorname{Bord}^{\mathrm{fr}}_{n}) to be the free connective spectrum on a point, i.e. the sphere spectrum:

Proposition 3.1. There is a unique equivalence of spectra L(Bord⁡nfr)≅SL(\operatorname{Bord}^{\mathrm{fr}}_{n}) \cong S with the property that the positively framed point gets mapped to 1∈π0[S,S]≅π0(S)1 \in\pi_0[S,S] \cong\pi_0(S).

Proof. For any connective spectrum XX, we compute:

Hom⁡Sp⁡≥0(L(Bord⁡nfr),X)→(1)Hom⁡Cat⁡(∞,n)E∞(Bord⁡nfr,ι(X))→(2)ι(X)fd→(3)Ω∞X.(2)\operatorname{Hom}_{\operatorname{Sp}_{\geq0}}\left(L(\operatorname{Bord}^{\mathrm{fr}}_{n}),X\right) \xrightarrow{(1)} \operatorname{Hom}_{\operatorname{Cat}^{\mathrm{E}_{\infty}}_{(\infty,n)}}\left(\operatorname{Bord}^{\mathrm{fr}}_{n},\iota(X)\right) \xrightarrow{(2)} \iota(X)^{\mathrm{fd}} \xrightarrow{(3)} \Omega^\infty X. \tag*{(2)}

Here:

♮1 is the L⊣ιL \dashv\iota adjunction.

♮2 is the cobordism hypothesis (Theorem 2.6), which states that symmetric monoidal functors Bord⁡nfr→ι(X)\operatorname{Bord}^{\mathrm{fr}}_{n} \to\iota(X) correspond to fully dualisable objects of ι(X)\iota(X).

♮3 is the observation that ι(X)fd\iota(X)^{\mathrm{fd}}, the ∞\infty-groupoid of fully dualisable objects of ι(X)\iota(X) is equivalent to the underlying space Ω∞X\Omega^\infty X. Indeed, every object of ι(X)\iota(X) is fully dualisable, so the core is exactly the space of objects.

The entire chain is natural in XX. Thus L(Bord⁡nfr)L(\operatorname{Bord}^{\mathrm{fr}}_{n}) represents the ∞\infty-groupoid-valued functor X↦Ω∞XX \mapsto\Omega^\infty X on Sp⁡≥0\operatorname{Sp}_{\geq0}. The sphere spectrum SS satisfies the same universal property, since Hom⁡Sp⁡(S,X)≃Ω∞X\operatorname{Hom}_{\operatorname{Sp}}(S,X) \simeq\Omega^\infty X for any spectrum XX. By the ∞\infty-categorical Yoneda lemma, we obtain an equivalence L(Bord⁡nfr)≃SL(\operatorname{Bord}^{\mathrm{fr}}_{n}) \simeq S.

Explicitly, the equivalence is determined by the image of the generating object xx: the unique map L(Bord⁡nfr)→SL(\operatorname{Bord}^{\mathrm{fr}}_{n}) \to S sending xx to 1∈π0(S)⊂Ω∞S1 \in\pi_0(S) \subset\Omega^\infty S is an equivalence. Any other equivalence L(Bord⁡nfr)≃SL(\operatorname{Bord}^{\mathrm{fr}}_{n}) \simeq S would differ by an automorphism of SS. Since π0Aut⁡(S)=π0[S,S]×=Z×={±1}\pi_0\operatorname{Aut}(S)=\pi_0[S,S]^\times=\mathbb{Z}^\times=\{\pm1\}, this would change L(Bord⁡nfr)→SL(\operatorname{Bord}^{\mathrm{fr}}_{n}) \to S by mapping the positively framed point to −1∈S-1 \in S. □

Remark 3.2. A similar argument appears in [12 Section 7].

Remark 3.3. Since SS is the monoidal unit in Sp≥0\mathrm{Sp}_{\geq0}, it is the unique spectrum so that Hom⁡Sp≥0(S,X)≃X\operatorname{Hom}_{\mathrm{Sp}_{\geq0}}(S,X)\simeq X for all XX; any other such spectrum YY would be right adjoint to Y∧(−)Y\wedge(-). We thus obtain the equivalence

L(Bordnfr)≃SL(\mathrm{Bord}^{\mathrm{fr}}_n)\simeq S

of the above proposition also from uniqueness of adjoint functors.

As a consequence, the equivalence is given by

L(Bordnfr)⟶L(Bordnfr)∧Hom⁡Sp≥0(L(Bordnfr),S)→evSL(\mathrm{Bord}^{\mathrm{fr}}_n)\longrightarrow L(\mathrm{Bord}^{\mathrm{fr}}_n)\wedge\operatorname{Hom}_{\mathrm{Sp}_{\geq0}}(L(\mathrm{Bord}^{\mathrm{fr}}_n),S)\xrightarrow{\mathrm{ev}} S

where the first map is the inverse of (2) for X=SX=S. In other words, it evaluates a framed cobordism MM on the unique TQFT Bordn→S\mathrm{Bord}_n\to S that assigns 11 to the positively framed point.

The inverse map S→L(Bordnfr)S\to L(\mathrm{Bord}^{\mathrm{fr}}_n) is given by the composition

S⟶S∧Hom⁡Sp(L(Bordnfr),L(Bordnfr))⟶Hom⁡Sp(L(Bordnfr),L(Bordnfr))⟶L(Bordnfr)S\longrightarrow S\wedge\operatorname{Hom}_{\mathrm{Sp}}(L(\mathrm{Bord}^{\mathrm{fr}}_n),L(\mathrm{Bord}^{\mathrm{fr}}_n))\longrightarrow\operatorname{Hom}_{\mathrm{Sp}}(L(\mathrm{Bord}^{\mathrm{fr}}_n),L(\mathrm{Bord}^{\mathrm{fr}}_n))\longrightarrow L(\mathrm{Bord}^{\mathrm{fr}}_n)

where the first map is induced by the identity on L(Bordnfr)L(\mathrm{Bord}^{\mathrm{fr}}_n), the second map is the canonical SS-module structure that exists on any spectrum, and the final map is (2) applied to X=L(Bordnfr)X=L(\mathrm{Bord}^{\mathrm{fr}}_n).

Tangential versus normal framings

Definition 3.4. Let ξ:X→BO(n)\xi:X\to BO(n) be a rank-nn vector bundle over XX.

♮1 An unstable framing of ξ\xi is a nullhomotopy of ξ\xi, i.e. an isomorphism of vector bundles Rn×X≅ξ\mathbb{R}^n\times X\cong\xi.

♮2 A stable framing of ξ\xi is a nullhomotopy of the composition X→ξBO(n)→BOX\xrightarrow{\xi} BO(n)\to BO.

A stable (resp. unstable) framing of a manifold (with corners) MM is a stable (resp. unstable) framing of its tangent bundle τM\tau_M.

If XX is compact, a stable framing is equivalent to a trivialisation of the bundle ξ⊕RN\xi\oplus\mathbb{R}^N for NN sufficiently large. Since a stably trivial bundle of rank >d>d over a kk-dimensional CW-complex is trivial, it suffices to take N=1N=1 if MM is a manifold and ξ=τM\xi=\tau_M. However, the space of stable framings is a torsor over the ∞\infty-group ΩMap⁡(M,BO)≃Map⁡(M,O)\Omega\operatorname{Map}(M,BO)\simeq\operatorname{Map}(M,O), which is in general different from Map⁡(M,O(n+1))\operatorname{Map}(M,O(n+1)).

If j:Mn↪Nn+kj:M^n\hookrightarrow N^{n+k} is an embedding between manifolds, the kk-dimensional quotient bundle νj:=j∗τN/τM\nu_j:=j^*\tau_N/\tau_M is called the normal bundle. Since every exact sequence of vector bundles splits noncanonically, in case N=Rn+kN=\mathbb{R}^{n+k}, this picks out a bundle νj\nu_j such that νj⊕τM≅Rn+k\nu_j\oplus\tau_M\cong\mathbb{R}^{n+k}. By the Whitney embedding theorem, such jj always exists. Moreover, the induced stable normal bundle M→BO(k)→BOM\to BO(k)\to BO is independent of the choice of jj and will be denoted νM\nu_M. This stable bundle is a choice of inverse of the stable tangent bundle under ⊕\oplus.

Since ⊖τM=νM\ominus\tau_M=\nu_M, a stable framing is equivalent to a trivialisation of the stable normal bundle, but the OO-actions differ by ⊖\ominus. More geometrically, suppose we have a given j:M→Rn+kj:M\to\mathbb{R}^{n+k} giving, after a contractible choice of metric, a trivialisation νj⊕τM≅Rn+k\nu_j\oplus\tau_M\cong\mathbb{R}^{n+k}. Then composing a given trivialisation τM≅Rn\tau_M\cong\mathbb{R}^n with g∈O(n)g\in O(n) corresponds to composing νM≅Rk\nu_M\cong\mathbb{R}^k with some g′∈O(k)g'\in O(k) and a homotopy between

νj⊕τM≅Rn+k→g⊕g′Rn+k\nu_j\oplus\tau_M\cong\mathbb{R}^{n+k}\xrightarrow{g\oplus g'}\mathbb{R}^{n+k}

and the given trivialisation of νj⊕τM\nu_j\oplus\tau_M, i.e. a nullhomotopy of g⊕g′g\oplus g'. We see that we need to use the inverse OO-action. This distinction will be crucial below: with the right-action convention used in this paper, the tangential action is ⊖Jn\ominus J_n, while the normal action is JnJ_n.

Remark 3.5. The subtlety about different OO-actions is relevant since the map ⊖:BO→BO\ominus: BO \to BO is not nullhomotopic. For example, consider the truncation t:BO→BO/BSpin≃BZ/2×BZ/2t: BO \to BO/BSpin \simeq B\mathbb{Z}/2 \times B\mathbb{Z}/2 explicitly given by (w1,w2)(w_1,w_2). Then by the Whitney sum formula, composition of ⊖\ominus with tt results in the map (w1,w2+w12)(w_1,w_2+w_1^2).

The abstract Pontryagin–Thom collapse map

We use the abstract formulation of the Pontryagin–Thom collapse map in terms of Spanier–Whitehead duality, following [2 Definition 10.3 and Proposition 10.5]. Let SS denote the sphere spectrum.

Definition 3.6. Let XX be a compact dd-manifold, and fix an embedding j:X↪Rd+kj:X \hookrightarrow\mathbb{R}^{d+k} with tubular neighbourhood NXN_X and normal bundle νj\nu_j. The classical Pontryagin–Thom collapse map is the based map

cj:Sd+k⟶Th⁡(νj)c_j:S^{d+k} \longrightarrow\operatorname{Th}(\nu_j)

obtained by collapsing the complement of NXN_X in Rd+k\mathbb{R}^{d+k} (compactified to Sd+kS^{d+k}) to the basepoint of the Thom space Th⁡(νj)\operatorname{Th}(\nu_j).

Passing to suspension spectra and using the canonical identification Th⁡(νj)≃Σ−d(X−TX)\operatorname{Th}(\nu_j) \simeq\Sigma^{-d}(X^{-T X}) (which holds because νj⊕τX≅Rd+k\nu_j \oplus\tau_X \cong\mathbb{R}^{d+k}), this gives the Pontryagin–Thom collapse map

cX:S⟶X−TX,c_X:S \longrightarrow X^{-T X},

where X−TXX^{-T X} denotes the Thom spectrum of the virtual bundle −TX-T X (the negative tangent bundle of XX), i.e. the colimit

X−TX:=colim⁡(X→−TXBO×Z→JPic⁡(S)→Sp⁡).X^{-T X}:=\operatorname{colim}\left(X \xrightarrow{-T X} BO \times\mathbb{Z} \xrightarrow{J} \operatorname{Pic}(S) \to\operatorname{Sp}\right).

The homotopy class of cXc_X is independent of the choice of embedding jj and tubular neighbourhood NXN_X: any two choices are related by an isotopy, and the space of tubular neighbourhoods of a fixed embedding is contractible.

Proposition 3.7. Let MM be a closed dd-manifold with a framing ϕ:TM≅Rd\phi:TM \cong\mathbb{R}^d. Then:

  1. The framing ϕ\phi induces an equivalence M−TM≃Σ−dΣ+∞MM^{-T M} \simeq\Sigma^{-d}\Sigma^\infty_+ M.

  2. The composition S→cMM−TM→ϕΣ−dΣ+∞M→crush⁡Σ−dSS \xrightarrow{c_M} M^{-T M} \xrightarrow{\phi} \Sigma^{-d}\Sigma^\infty_+ M \xrightarrow{\operatorname{crush}} \Sigma^{-d}S defines an element of πdS\pi_d S.

  3. This assignment is the classical Pontryagin–Thom isomorphism: it identifies the framed bordism group Ωdfr\Omega_d^{\mathrm{fr}} with πdS\pi_d S.

Proof. (1) follows from the fact that −TM≅−Rd-TM \cong-\mathbb{R}^d under the framing ϕ\phi, and the Thom spectrum of a trivial bundle of rank −d-d is Σ−dΣ+∞M\Sigma^{-d}\Sigma^\infty_+M. (2) is the definition of the map: cMc_M is the abstract PT collapse map of Definition 3.6, ϕ\phi is the trivialisation induced by the framing, and the crush map is induced by M→ptM \to\mathrm{pt}. The fact that it is an isomorphism is the classical Pontryagin–Thom theorem [18 Chapter IX]. □

Lemma 3.8. For g∈O(d)g \in O(d), let g⋅ϕ:=ϕ∘gg \cdot\phi:= \phi\circ g denote the framing obtained by acting on ϕ\phi by the right action of O(d)O(d). Then the map

S⟶Σ−dSS \longrightarrow\Sigma^{-d}S

constructed in Proposition 3.7 for the framing g⋅ϕg \cdot\phi is obtained from the map for ϕ\phi by post-composing with the one-point compactification of g−1:Rd→Rdg^{-1}:\mathbb{R}^d \to\mathbb{R}^d (acting on the target Σ−dS=S−d\Sigma^{-d}S=S^{-d}). The resulting element of πdS\pi_dS is Jd(g)=Jd(g)−1J_d(g)=J_d(g)^{-1}.

Proof. Changing the framing by gg changes the induced trivialisation of −TM-TM by g−1g^{-1}. After the PT construction and crushing M+→S0M_+ \to S^0, the induced map on the target Σ−dS=S−d\Sigma^{-d}S=S^{-d} is the one-point compactification of g−1:Rd→Rdg^{-1}:\mathbb{R}^d \to\mathbb{R}^d, whose stable class is Jd(g)−1=⊖Jd(g)J_d(g)^{-1}=\ominus J_d(g). □

Theorem 3.9. The Pontryagin–Thom construction induces an equivalence of spectra

PT:L(Bordnfr)→∼S.\mathrm{PT}: L(\mathrm{Bord}^{\mathrm{fr}}_n) \xrightarrow{\sim} S.

Proof. The assignment (M,ϕ)↦cM,ϕ(M,\phi) \mapsto c_{M,\phi} of Proposition 3.7 is compatible with disjoint union, so it defines a symmetric monoidal functor PT:Bordnfr→ι(S)\mathrm{PT}: \mathrm{Bord}^{\mathrm{fr}}_n \to\iota(S). Passing to localisations gives a map PT:L(Bordnfr)→S\mathrm{PT}: L(\mathrm{Bord}^{\mathrm{fr}}_n) \to S.

Step 1: Isomorphism on π0\pi_0. The positively framed point maps to 1∈π0S1 \in\pi_0S.

Step 2: Isomorphism on πk\pi_k for k<n−1k < n - 1. By Corollary 3.13, πkL(Bordnfr)≅Ωkfr\pi_kL(\mathrm{Bord}^{\mathrm{fr}}_n) \cong\Omega_k^{\mathrm{fr}}. The classical Pontryagin–Thom theorem identifies Ωkfr≅πkS\Omega_k^{\mathrm{fr}} \cong\pi_kS. For k<n−1k < n - 1, an nn-framing on a kk-manifold is equivalent to a stable framing, so the unstable and stable PT constructions agree on πk\pi_k.

Step 3: Isomorphism on πk\pi_k for k≥n−1k \ge n - 1. Pass to the colimit over nn. By Lemma 3.16, πkL(Bord∞fr)≅Ωkfr≅πkS\pi_kL(\mathrm{Bord}^{\mathrm{fr}}_\infty) \cong\Omega_k^{\mathrm{fr}} \cong\pi_kS.

Step 4: Equivalence. Since both sides are connective spectra and PT is an isomorphism on all homotopy groups, it is an equivalence. □

Pontryagin–Thom: identifying L(Bordnfr)L(\mathrm{Bord}^{\mathrm{fr}}_n) with bordism groups

The Pontryagin–Thom theorem gives a geometric interpretation of the homotopy groups of the sphere spectrum in terms of bordism groups of stably framed manifolds. On the other hand, L(Bordnfr)L(\mathrm{Bord}^{\mathrm{fr}}_n) gives a geometric interpretation in terms of a certain spectrum built out of unstably framed manifolds. The goal of this section is to compare these descriptions and see how the Pontryagin–Thom theorem applies.

Given a symmetric monoidal (∞,n)(\infty,n)-category CC, define the (∞,n−1)(\infty,n-1)-category ΩC:=End⁡C1\Omega C := \operatorname{End}_C 1. There is a map ΩkC→ΩkL(C)\Omega^k C \to\Omega^k L(C) given by Ωk\Omega^k applied to the localisation map. By the universal property of LL this gives a map L(ΩkC)→ΩkL(C)L(\Omega^k C) \to\Omega^k L(C). For a general category, this map need not be close to an equivalence. However, if CC has duals, then localisation identifies those as inverses, and consequently we do get:

Lemma 3.10. If k<n−1k < n - 1, the map L(ΩkC)→ΩkL(C)L(\Omega^k C) \to\Omega^k L(C) is an isomorphism on π0\pi_0. Explicitly, we obtain an isomorphism of groups πkLC=ΩkC/∼\pi_kLC = \Omega^kC/\sim, where x∼yx \sim y if and only if xx and yy are connected by a k+1k+1-morphism.

Proof. Since k<nk < n, we can use the existence of adjoints of kk-morphisms to ensure that x,y∈ΩkCx,y \in\Omega^kC as opposed to a zigzag of kk-morphisms. There is no need to group-complete since ev⁡:x∨⊗x→1\operatorname{ev}: x^\vee\otimes x \to1 will ensure x∨x^\vee is the inverse of xx. Observe that ∼\sim is already an equivalence relation for k<n−1k < n - 1. □

For C=BordnfrC = \mathrm{Bord}^{\mathrm{fr}}_n, this means explicitly that πkLBordnfr\pi_kL\mathrm{Bord}^{\mathrm{fr}}_n for k<nk < n are given by bordism groups. Indeed, unpacking the definition of ΩkBordnfr\Omega^k\mathrm{Bord}^{\mathrm{fr}}_n we see that it has objects kk-dimensional closed manifolds YY with a trivialisation TY⊕Rn−k≅RnTY \oplus\mathbb{R}^{n-k} \cong\mathbb{R}^n, 11-morphisms k+1k+1-dimensional bordisms MM with a trivialisation TM⊕Rn−k−1≅RnTM \oplus\mathbb{R}^{n-k-1} \cong\mathbb{R}^n, et cetera.

Definition 3.11. Let B→BO(n)B \to BO(n) be a tangential structure and k<nk < n. The bordism group of kk-dimensional manifolds with BB-structure is the monoid of closed manifolds with BB-structure under disjoint union modulo k+1k+1-dimensional bordisms with BB-structure.

Remark 3.12. ∼\sim need not be an equivalence relation for k=n−1k = n - 1. For example, consider the case n=2n = 2 in Bord2fr\mathrm{Bord}^{\mathrm{fr}}_2. The manifold with a circle boundary obtained by puncturing a surface of genus g≥2g \ge2 has an unstable framing, since it is homotopy equivalent to a 1-dimensional CW-complex. We obtain a 1-morphism in Bord2fr\mathrm{Bord}^{\mathrm{fr}}_2 from S1S^1 with some framing of TS1⊕RTS^1 \oplus\mathbb{R} to ∅\varnothing. There cannot be a 1-morphism from ∅\varnothing to S1S^1 with the same framing, as their composition would result in an unstably framed closed manifold of genus g′≥g≥2g' \ge g \ge2.

However, for k<n−1k<n-1, adjoints also exist for kk-morphisms. Therefore, a morphism x→yx\to y exists if and only if a morphism y→xy\to x exists, and so ∼\sim is an equivalence relation. In particular,

Corollary 3.13. If k<nk<n, then πkL(Bordnfr)\pi_kL(\mathrm{Bord}^{\mathrm{fr}}_n) is isomorphic to the bordism group of unstably framed manifolds. If k<n−1k<n-1, it is isomorphic to the bordism group of stably framed manifolds.

Proof. The second point follows because in that case the k+1k+1-morphisms come with a once stabilised trivialisation of the tangent bundle, which is equivalent to a stable trivialisation. □

Since we only understand πkL(Bordnfr)\pi_kL(\mathrm{Bord}^{\mathrm{fr}}_n) in terms of bordism groups for small kk, we take n→∞n\to\infty. We comment briefly on the more mysterious geometric interpretations of πkL(Bordnfr)\pi_kL(\mathrm{Bord}^{\mathrm{fr}}_n) for k≥nk\geq n.

Example 3.14. For n=0n=0, the symmetric monoidal groupoid Bordnfr\mathrm{Bord}^{\mathrm{fr}}_n of zero-dimensional manifolds and diffeomorphisms is equivalent to the groupoid

∐k≥0BSk\coprod_{k\geq0} BS_k

of finite sets equipped with a disjoint union. The fact that

L(Bord0fr)≃SL(\mathrm{Bord}^{\mathrm{fr}}_0)\simeq S

is a version of the Barratt-Priddy-Quillen theorem.

Example 3.15. The group πnL(Bordnfr)\pi_nL(\mathrm{Bord}^{\mathrm{fr}}_n) is generated by zigzags of nn-dimensional framed bordisms

∅→M1Y1←M2Y2→M3⋯←MkYk→Mk+1∅\varnothing\xrightarrow{M_1} Y_1 \xleftarrow{M_2} Y_2 \xrightarrow{M_3}\cdots\xleftarrow{M_k} Y_k \xrightarrow{M_{k+1}} \varnothing

between (n−1)(n-1)-dimensional compact manifolds with a trivialisation TYj⊕R≅RnTY_j\oplus\mathbb{R}\cong\mathbb{R}^n. The isomorphism of groups πnL(Bordnfr)→πnL(Bordn+1fr)≅Ωnfr\pi_nL(\mathrm{Bord}^{\mathrm{fr}}_n)\to\pi_nL(\mathrm{Bord}^{\mathrm{fr}}_{n+1})\cong\Omega_n^{\mathrm{fr}} is explicitly given by taking the composition to a stably framed bordism ∅→∅\varnothing\to\varnothing after potentially reversing bordisms by taking adjoints.

There is a symmetric monoidal functor Bordnfr→Bordn+1fr\mathrm{Bord}^{\mathrm{fr}}_n\to\mathrm{Bord}^{\mathrm{fr}}_{n+1} that maps a kk-manifold YY for k≤nk\leq n with TY⊕Rn−k≅RnTY\oplus\mathbb{R}^{n-k}\cong\mathbb{R}^n to the same manifold with the isomorphism TY⊕Rn+1−k≅Rn+1TY\oplus\mathbb{R}^{n+1-k}\cong\mathbb{R}^{n+1} obtained by stabilising with the identity, and a diffeomorphism of nn-dimensional manifolds to its mapping cylinder as an n+1n+1-dimensional manifold with the obvious unstable framing. This functor is clearly O(n)O(n)-equivariant when restricting the O(n+1)O(n+1)-action on Bordn+1fr\mathrm{Bord}^{\mathrm{fr}}_{n+1} to the first nn coordinates. We thus get an OO-action on

colim⁡nBordnfr∈Cat∞E∞\operatorname*{colim}_{n}\mathrm{Bord}^{\mathrm{fr}}_n\in\mathrm{Cat}^{E_\infty}_\infty

where Cat∞E∞=colim⁡nCat(∞,n)E∞\mathrm{Cat}^{E_\infty}_\infty=\operatorname*{colim}_{n}\mathrm{Cat}^{E_\infty}_{(\infty,n)} is the (∞,1)(\infty,1)-category of (∞,∞)(\infty,\infty)-categories defined as a colimit in PrL\mathrm{Pr}^{L}. We will be interested in the induced OO-action on L(Bord∞fr)≃colim⁡nL(Bordnfr)L(\mathrm{Bord}^{\mathrm{fr}}_\infty)\simeq\operatorname*{colim}_{n}L(\mathrm{Bord}^{\mathrm{fr}}_n).

Lemma 3.16. The degree kk homotopy group of the spectrum L(Bord∞fr)L(\mathrm{Bord}^{\mathrm{fr}}_\infty) is given by Ωkfr\Omega_k^{\mathrm{fr}}, the bordism group of stably framed kk-dimensional manifolds.

Proof. The functor πk:S∗→Ab\pi_k:\mathcal{S}_*\to\mathrm{Ab} preserves filtered colimits because SkS^k is compact, and so

πkL(colim⁡nBordnfr)=colim⁡nπkL(Bordnfr).\pi_kL\left(\operatorname*{colim}_{n}\mathrm{Bord}^{\mathrm{fr}}_n\right)=\operatorname*{colim}_{n}\pi_kL(\mathrm{Bord}^{\mathrm{fr}}_n).

For nn large enough, πkL(Bordnfr)≅Ωkfr\pi_kL(\mathrm{Bord}^{\mathrm{fr}}_n)\cong\Omega_k^{\mathrm{fr}} and the structure maps in the colimit are the identity. □

:::

In Section 3.1, we concluded that there are equivalences L(Bord⁡nfr)≃SL(\operatorname{Bord}^{\mathrm{fr}}_n) \simeq S for all nn. The map Bord⁡nfr→Bord⁡n+1fr\operatorname{Bord}^{\mathrm{fr}}_n \to\operatorname{Bord}^{\mathrm{fr}}_{n+1} induces the identity map on SS after applying LL. Indeed, since the cobordism hypothesis is given by evaluating on a point, the diagram

Fun⁡(Bord⁡nfr,X)→Fun⁡(Bord⁡n+1fr,X)↓↓X←X\begin{CD} \operatorname{Fun}(\operatorname{Bord}^{\mathrm{fr}}_n,X) @>>> \operatorname{Fun}(\operatorname{Bord}^{\mathrm{fr}}_{n+1},X) \\ @VVV @VVV \\ X @<<< X \end{CD}

commutes. We therefore also get an equivalence colim⁡nL(Bord⁡nfr)≃S\operatorname*{colim}_{n} L(\operatorname{Bord}^{\mathrm{fr}}_n) \simeq S.

Recall from Definition 3.6 that the Pontryagin–Thom collapse map sends a closed normally framed dd-manifold Y↪RNY \hookrightarrow\mathbb{R}^{N} to a based map SN→SN−dS^{N} \to S^{N-d}, whose stable homotopy class lies in πdS\pi_d S.

Theorem 3.17. On homotopy groups, the equivalence L(Bord⁡∞fr)≃SL(\operatorname{Bord}^{\mathrm{fr}}_{\infty}) \simeq S is given by the Pontryagin–Thom construction.

Proof. (Sketch) It suffices to show that the Pontryagin–Thom construction defines an equivalence of spectra L(Bord⁡∞fr)≃SL(\operatorname{Bord}^{\mathrm{fr}}_{\infty}) \simeq S. Indeed, we know that SS has only one nontrivial automorphism, and the Pontryagin–Thom construction sends ++ to 1∈π0S1 \in\pi_0 S. We therefore want to prove that the Pontryagin–Thom construction defines compatible nn-dimensional TQFTs Bord⁡nfr≃S\operatorname{Bord}^{\mathrm{fr}}_n \simeq S.

We sketch a scanning argument for a model of Bord⁡nB\operatorname{Bord}^{B}_n in which bordisms are embedded in R∞\mathbb{R}^{\infty} with an nn-dimensional hypercube picked out for composition, and another infinitely many coordinates for the E∞E_{\infty}-structure. We will take BB to be the (stable) tangential structure of a (stable) normal framing; Bord⁡nfr\operatorname{Bord}^{\mathrm{fr}}_n clearly maps to this, and this is all we need for the PT construction.

So suppose XX is an nn-dimensional bordism embedded in RN×[0,1]\mathbb{R}^{N} \times[0,1] from an (n−1)(n-1)-dimensional closed manifold Y0Y_0 embedded in RN×{0}\mathbb{R}^{N} \times\{0\} to an (n−1)(n-1)-dimensional closed manifold Y1Y_1 embedded in RN×{1}\mathbb{R}^{N} \times\{1\}. The Thom–Pontryagin construction then gives a based map SN×[0,1]→SN−nS^{N} \times[0,1] \to S^{N-n} which restricts to the Thom–Pontryagin construction of Y1Y_1 and Y2Y_2 on the boundaries, which is the desired morphism in [SN,SN−n][S^{N},S^{N-n}]. Observe that bordisms compose as homotopies do. A similar construction holds for an arbitrary kk-morphism contained in [0,1]k×RN[0,1]^k \times\mathbb{R}^{N} giving a map SN×[0,1]k→SN−n−kS^{N} \times[0,1]^k \to S^{N-n-k}. The space of tubular neighbourhoods is contractible, so the construction is well-defined up to coherent homotopy. □

Identifying the O(n)O(n)-action

Theorem 3.18 (Main theorem). With respect to the right action of O(n)O(n) on framings (Convention 1), the O(n)O(n)-action on L(Bord⁡nfr)≃SL(\operatorname{Bord}^{\mathrm{fr}}_n) \simeq S induced by change of tangential framing is given by

⊖Jn ⁣:O(n)⟶GL1(S),g↦Jn(g)−1.\ominus J_n\colon O(n) \longrightarrow\mathrm{GL}_1(S), \qquad g \mapsto J_n(g)^{-1}.

Passing to the colimit as n→∞n \to\infty, the OO-action on colim⁡nL(Bord⁡nfr)≃S\operatorname*{colim}_{n}L(\operatorname{Bord}^{\mathrm{fr}}_n) \simeq S is given by ⊖J ⁣:O→GL1(S)\ominus J\colon O \to\mathrm{GL}_1(S).

Proof. By Theorem 3.9, the Pontryagin–Thom construction gives an equivalence of spectra PT ⁣:L(Bord⁡nfr)→S\mathrm{PT}\colon L(\operatorname{Bord}^{\mathrm{fr}}_n) \to S, under which the positively framed point corresponds to 1∈π0S1 \in\pi_0 S. By Lemma 3.8, this equivalence is O(n)O(n)-equivariant, with the action of gg on SS given by ⊖Jn(g)=Jn(g)−1\ominus J_n(g) = J_n(g)^{-1}.

More explicitly, on the standard nn-ball Bn⊂RnB^n \subset\mathbb{R}^n with its standard framing, the PT map is the identity Sn→SnS^n \to S^n (as an element of πnS\pi_n S). Changing the framing by gg (right action) rotates the target sphere by g−1g^{-1}, giving ⊖Jn(g)∈GL1(S)\ominus J_n(g) \in\mathrm{GL}_1(S). Since the nn-balls generate SS as a spectrum, the action on all of SS is ⊖Jn(g)\ominus J_n(g).

Passing to the colimit as n→∞n \to\infty gives the stable JJ-homomorphism composed with ⊖ ⁣:⊖J ⁣:O→GL1(S)\ominus\colon\ominus J\colon O \to\mathrm{GL}_1(S). □

Remark 3.19. The O(n)O(n)-action can be seen directly in the classical Pontryagin–Thom construction. Embed Md↪Rd+kM^d \hookrightarrow\mathbb{R}^{d+k} with a normal framing. Changing the tangential framing of MM by g∈O(d)g \in O(d) is equivalent to rotating the embedding coordinates by g⊕id⁡k∈O(d+k)g \oplus\operatorname{id}_k \in O(d+k), which acts on the source sphere Sd+kS^{d+k} in the collapse map Sd+k→SkS^{d+k} \to S^k. After stabilisation, this rotation is the unstable JJ-homomorphism Jd(g)J_d(g). With the right-action convention for framings, the normal framing changes by g−1g^{-1}, so the collapse map is post-composed with Jd(g)−1J_d(g)^{-1}, giving ⊖Jd(g)\ominus J_d(g) (Theorem 3.18).

Remark 3.20. The idea that the O(n)O(n)-action on the space of fully dualisable objects in a symmetric monoidal (∞,n)(\infty,n)-category with duals is related to the JJ-homomorphism that appears in Lurie’s original treatment of the cobordism hypothesis [20 Example 2.4.15]. Lurie sketches that for a Picard ∞\infty-groupoid, the O(n)O(n)-action is “a version of the classical JJ-homomorphism”. In this paper we make this identification explicit: using the right-action convention for the O(n)O(n)-action on framings, we identify the action as ⊖Jn\ominus J_n.

Remark 3.21. An equivalent result has been obtained by Grady–Pavlov [16 Theorem 5.0.6] using the framework of geometric cobordism categories. The proof above works directly within Lurie’s formulation of the cobordism hypothesis, using the abstract Pontryagin–Thom collapse map. Our convention (right action of O(n)O(n) on framings) gives the action ⊖Jn\ominus J_n; in the left-action convention, the same proof gives JnJ_n.

The classification of invertible field theories

An invertible TQFT is a TQFT in which the target has invertible kk-morphisms for all k≥0k \ge0. More precisely, an nn-dimensional BB-structured invertible TQFT with target a connective spectrum XX is a symmetric monoidal functor Bord⁡nB→ιX\operatorname{Bord}^{B}_{n} \to\iota X, or equivalently, a map of spectra LBord⁡nB→XL\operatorname{Bord}^{B}_{n} \to X. The following result is at the cornerstone of modern computations of anomalies and invertible phases of matter [11], and was shown using different techniques by [26].

Proposition 3.22. If ξ:B→BO(n)\xi: B \to BO(n) is a tangential structure, there is an equivalence of spectra LBord⁡nB≃MTBL\operatorname{Bord}^{B}_{n} \simeq MTB natural in BB.

Proof. We use the equivalence Bord⁡nB≃colim⁡BBord⁡nfr\operatorname{Bord}^{B}_{n} \simeq\operatorname*{colim}_{B}\operatorname{Bord}^{\mathrm{fr}}_{n}. Since LL preserves colimits, we get

L(Bord⁡nB)≃colim⁡BL(Bord⁡nfr)≃colim⁡BS,L(\operatorname{Bord}^{B}_{n}) \simeq\operatorname*{colim}_{B} L(\operatorname{Bord}^{\mathrm{fr}}_{n}) \simeq\operatorname*{colim}_{B} S,

where the BB-action on SS is via the composition

B→ξBO(n)→(ι,n)BO×Z→⊖BO×Z→JPic⁡(S)→Sp⁡,B \xrightarrow{\xi} BO(n) \xrightarrow{(\iota,n)} BO \times\mathbb{Z} \xrightarrow{\ominus} BO \times\mathbb{Z} \xrightarrow{J} \operatorname{Pic}(S) \to\operatorname{Sp},

using Theorem 3.18 (with the right-action convention, the O(n)O(n)-action is ⊖Jn\ominus J_n, and passing to BB gives the composition above). By Definition 2.16, this colimit is exactly the Madsen–Tillmann spectrum MTBMTB (the Thom spectrum of the virtual normal bundle −ξ-\xi). Therefore

L(Bord⁡nB)≃MTB.L(\operatorname{Bord}^{B}_{n}) \simeq MTB.

Naturality in BB follows from the naturality in each step. □

Remark 3.23. For the classification of invertible TFTs, the identification L(Bord⁡nB)≃MTBL(\operatorname{Bord}^{B}_{n}) \simeq MTB means that a BB-structured invertible TFT is a map MTB→XMTB \to X of spectra, i.e., an element of π0Map⁡Sp⁡(MTB,X)\pi_0\operatorname{Map}_{\operatorname{Sp}}(MTB,X). Equivalently, it is a homotopy fixed point XhBX^{hB}, where BB acts on XX via ⊖Jn\ominus J_n (Theorem 3.25).

Remark 3.24. With the right-action convention for the O(n)O(n)-action on framings (Convention 1), the O(n)O(n)-action on L(Bord⁡nfr)L(\operatorname{Bord}^{\mathrm{fr}}_{n}) is ⊖Jn\ominus J_n, not JnJ_n (Theorem 3.18). This is because the change of framing acts by g−1g^{-1} on the induced trivialisation of the tangent bundle. Equivalently, one can use the left-action convention for the O(n)O(n)-action on framings, in which case the action is JnJ_n; the two conventions differ by the inversion automorphism ⊖\ominus (Convention 1). In this paper we use the right-action convention, so the action is ⊖Jn\ominus J_n.

Now, it follows that

Theorem 3.25. The spectrum of BB-structured invertible topological field theories valued in a connective spectrum XX is equivalent to

Map⁡Sp(L(Bord⁡nB),X)≃Map⁡Sp(MTB,X),\operatorname{Map}_{\mathrm{Sp}}(L(\operatorname{Bord}^{B}_{n}),X) \simeq\operatorname{Map}_{\mathrm{Sp}}(MTB,X),

where the BB-action on XX is via the composition

B→ξBO(n)→B(⊖Jn)BGL1(S)→⊗XSp.B \xrightarrow{\xi} BO(n) \xrightarrow{B(\ominus J_n)} BGL_1(S) \xrightarrow{\otimes X} \mathrm{Sp}.

Equivalently, the spectrum is the homotopy fixed point spectrum XhBX^{hB}.

Proof. By Proposition 3.22, L(Bord⁡nB)≃MTBL(\operatorname{Bord}^{B}_{n}) \simeq MTB. The spectrum of BB-structured invertible TFTs valued in XX is therefore

Map⁡Sp(L(Bord⁡nB),X)≃Map⁡Sp(MTB,X).\operatorname{Map}_{\mathrm{Sp}}(L(\operatorname{Bord}^{B}_{n}),X) \simeq\operatorname{Map}_{\mathrm{Sp}}(MTB,X).

For B=ptB = \mathrm{pt}, we have L(Bord⁡nfr)≃SL(\operatorname{Bord}^{\mathrm{fr}}_{n}) \simeq S, so the spectrum is Map⁡Sp(S,X)≃X\operatorname{Map}_{\mathrm{Sp}}(S,X) \simeq X. The O(n)O(n)-action is induced by precomposition with the ⊖Jn\ominus J_n-action on SS. By adjunction, this is the action on XX via the SS-module structure. Taking homotopy fixed points yields XhBX^{hB}. □

Remark 3.26 (Conjectural). Given a connective commutative ring spectrum RR, define Mod⁡(n)(R)\operatorname{Mod}^{(n)}(R) inductively as the higher idempotent completion [15] of the delooping of the symmetric monoidal (∞,n−1)(\infty,n-1)-category Mod⁡(n−1)(R)\operatorname{Mod}^{(n-1)}(R), where Mod⁡(0)(R):=R\operatorname{Mod}^{(0)}(R) := R. Let Pic⁡(n)(R)\operatorname{Pic}^{(n)}(R) be the maximal Picard sub-∞\infty-groupoid of Mod⁡(n)(R)\operatorname{Mod}^{(n)}(R). Then Pic⁡(2)(R)\operatorname{Pic}^{(2)}(R) is the Brauer–Picard spectrum.

For R=SR = S, Pic⁡(2)(S)\operatorname{Pic}^{(2)}(S) is known to be BPic⁡(S)B\operatorname{Pic}(S) [5]. As far as we know, Pic⁡(n)(S)\operatorname{Pic}^{(n)}(S) is not known for n>2n > 2. The O(n)O(n)-action on this space will, by the cobordism hypothesis, be induced from the O(n)O(n)-action on Bord⁡nfr\operatorname{Bord}^{\mathrm{fr}}_{n}, whose localisation is computed by Theorem 3.18 to be ⊖Jn\ominus J_n.

Computations

In this section, we illustrate Theorem 3.18 by determining the O(n)O(n)-action on the homotopy groups of the sphere spectrum SS in low degrees. By Theorem 3.18, the O(n)O(n)-action on L(Bord⁡nfr)≃SL(\operatorname{Bord}^{\mathrm{fr}}_{n}) \simeq S is given by the composition

O(n)→⊖JnGL1(S)⟶Aut⁡(S),O(n) \xrightarrow{\ominus J_n} GL_1(S) \longrightarrow\operatorname{Aut}(S),

where the second map sends a unit to the corresponding multiplication map on SS. We determine the effect of this action on πkS\pi_k S for small kk and all nn.

The J-homomorphism concretely

The sphere spectrum SS plays the role of the integers in abelian groups; it is the initial E∞E_\infty-ring spectrum, and every spectrum XX has a canonical SS-module structure. If X∈SpX \in\mathrm{Sp}, the map S×X→XS \times X \to X induces action maps πkS×πlX→πk+lS\pi_k S \times\pi_l X \to\pi_{k+l}S.

Recall the homotopy groups of the sphere spectrum in low degrees [25]

π0S≅Z,π1S≅Z/2 (generated by η),π2S≅Z/2 (generated by η2),π3S≅Z/24 (generated by ν).\pi_0 S \cong\mathbb{Z}, \qquad\pi_1 S \cong\mathbb{Z}/2 \text{ (generated by } \eta\text{)}, \qquad\pi_2 S \cong\mathbb{Z}/2 \text{ (generated by } \eta^2\text{)}, \qquad\pi_3 S \cong\mathbb{Z}/24 \text{ (generated by } \nu\text{)}.

Lemma 4.1. For k=0k = 0, the action map is given by −1 ⁣:πlX→πlX-1 \colon\pi_l X \to\pi_l X.

For k>0k > 0, the action is much more mysterious. For example, for for k=1k = 1 the generator of π1S≅Z/2\pi_1 S \cong\mathbb{Z}/2 gives a map πlX→πl+1X\pi_l X \to\pi_{l+1}X which by bilinearity of the action map factors through a map ηX ⁣:πlX→πl+1X/2\eta_X \colon\pi_l X \to\pi_{l+1}X/2.

Lemma 4.2. The stable cohomology group π0Map⁡Sp(πlX,Σπl+1X)\pi_0\operatorname{Map}_{\mathrm{Sp}}(\pi_l X,\Sigma\pi_{l+1}X) is isomorphic to Hom⁡Ab(πlX,πl+1X/2)\operatorname{Hom}_{\mathrm{Ab}}(\pi_l X,\pi_{l+1}X/2). The element ηX,l\eta_{X,l} of this group is the kk-invariant of XX at the spot l,l+1l,l+1.

We are not aware of a general way to describe the action πlX→πl+3X/24\pi_l X \to\pi_{l+3}X/24 of ν\nu.

Definition 2.9 gives the pullback

GL1(S)→Ω∞S↓↓π0{±1}→Z\begin{CD} \mathrm{GL}_1(S) @>>> \Omega^\infty S \\ @VVV @VV{\pi_0}V \\ \{\pm1\} @>>> \mathbb{Z} \end{CD}

in the ∞\infty-category S\mathcal{S}. From the associated fibre sequence, one reads off its homotopy groups:

π0GL1(S)≅Z/2≅{±1},πkGL1(S)≅πkSfor all k≥1.\pi_0\mathrm{GL}_1(S) \cong\mathbb{Z}/2 \cong\{\pm1\}, \qquad\pi_k\mathrm{GL}_1(S) \cong\pi_kS \quad\text{for all } k \ge1.

The image of the JJ-homomorphism is understood better than the homotopy groups of GL1(S)\mathrm{GL}_1(S). In low degrees, we have [4]

Z≅π0(BO×Z)→≅π0Pic⁡(S)≅Z\mathbb{Z} \cong\pi_0(BO \times\mathbb{Z}) \xrightarrow{\cong} \pi_0\operatorname{Pic}(S) \cong\mathbb{Z}
Z/2≅π0O→≅π0GL1(S)≅Z/2\mathbb{Z}/2 \cong\pi_0O \xrightarrow{\cong} \pi_0\mathrm{GL}_1(S) \cong\mathbb{Z}/2
Z/2≅π1O→≅π1GL1(S)≅Z/2∋η\mathbb{Z}/2 \cong\pi_1O \xrightarrow{\cong} \pi_1\mathrm{GL}_1(S) \cong\mathbb{Z}/2 \ni\eta
0≅π2O→0π2GL1(S)≅Z/2∋η20 \cong\pi_2O \xrightarrow{0} \pi_2\mathrm{GL}_1(S) \cong\mathbb{Z}/2 \ni\eta^2
Z≅π3O↠π3GL1(S)≅Z/24∋ν\mathbb{Z} \cong\pi_3O \twoheadrightarrow\pi_3\mathrm{GL}_1(S) \cong\mathbb{Z}/24 \ni\nu
0≅π4O→0π4GL1(S)≅00 \cong\pi_4O \xrightarrow{0} \pi_4\mathrm{GL}_1(S) \cong0
0≅π5O→0π5GL1(S)≅00 \cong\pi_5O \xrightarrow{0} \pi_5\mathrm{GL}_1(S) \cong0
0≅π6O→0π6GL1(S)≅Z/2∋ν20 \cong\pi_6O \xrightarrow{0} \pi_6\mathrm{GL}_1(S) \cong\mathbb{Z}/2 \ni\nu^2
Z≅π7O↠π7GL1(S)≅Z/240\mathbb{Z} \cong\pi_7O \twoheadrightarrow\pi_7\mathrm{GL}_1(S) \cong\mathbb{Z}/240

We obtain:

Proposition 4.3. ♯1 The generator of π0O\pi_0O acts on XX by −1-1. ♯2 The generator of π1O\pi_1O acts on XX by η\eta as described in Lemma 4.2. ♯3 The generator of π3O\pi_3O acts on XX by ν\nu.

The full OO-action on XX contains even more interesting information about how the homotopy groups of BOBO fit together.

Remark 4.4. The computation in Proposition 4.3 determines the O(n)O(n)-action on each πkS\pi_kS as a self-map. The full homotopy fixed point spectrum ShBS^{hB} for a tangential structure B→BO(n)B \to BO(n) is computed by the homotopy fixed point spectral sequence

E2p,q=Hp(B;π−qS)⟹π−p−q(ShB).E_2^{p,q}=H^p(B;\pi_{-q}S)\Longrightarrow\pi_{-p-q}(S^{hB}).

where the coefficients π∗S\pi_*S carry the BB-action described above. This spectral sequence computes the homotopy groups of the spectrum of BB-structured invertible TQFTs valued in SS. The spectral sequence also involves higher coherences of the action: the fact that BB acts via the map B→BO(n)→ΘJnGL1(S)B \to BO(n) \xrightarrow{\Theta J_n} \mathrm{GL}_1(S) means that elements of π1(B)\pi_1(B) can act by multiplication by η∈π1S\eta\in\pi_1S, and elements of π3(B)\pi_3(B) by ν∈π3S\nu\in\pi_3S. These enter the spectral sequence through differentials and Massey products, but do not affect the action on individual homotopy groups as self-maps.

For n=2n = 2, a rotation with odd winding number has image η∈π1GL1(S)\eta\in\pi_{1}\mathrm{GL}_{1}(S). For n≥3n \ge3, the nontrivial element of π1(SO(n))\pi_{1}(SO(n)) maps to η\eta [4]. In both cases, this π1\pi_{1}-information contributes to the d2d_{2}-differential in the spectral sequence, encoding the fact that a 2π2\pi rotation acts nontrivially on the space of invertible TFTs, even though it acts trivially on the homotopy groups of SS as self-maps. This is the mathematical mechanism behind the spin-statistics theorem in the context of framed field theories [11].

There is a direct relationship between this spectral sequence and the Madsen–Tillmann spectrum MTBMTB. By the classification of invertible TFTs (Theorem 3.25), we have L(BordnB)≃MTBL(\mathrm{Bord}_{n}^{B}) \simeq MTB, and therefore

ShB≃Map⁡Sp(MTB,S).S^{hB} \simeq\operatorname{Map}_{\mathrm{Sp}}(MTB,S).

Consequently, the homotopy fixed point spectral sequence converges to the homotopy groups of the function spectrum Map⁡Sp(MTB,S)\operatorname{Map}_{\mathrm{Sp}}(MTB,S). For specific choices of BB, the differentials in this spectral sequence encode the bordism groups of BB-structured manifolds: the d2d_{2}-differential, governed by the η\eta-action, obstructs the existence of certain BB-structured invertible TFTs in dimension nn; higher differentials involving ν\nu obstruct the extension of such theories to dimension n+1n+1.

What these computations determine

In the previous subsections we computed the action for target SS; we now explain how this generalises to an arbitrary connective spectrum XX. By Corollary 2.8 applied to C=ι(X)C = \iota(X) (so that Cfd≃Ω∞XC^{\mathrm{fd}} \simeq\Omega^\infty X and the homotopy fixed points are taken at the spectrum level), for a tangential structure ξ ⁣:B→BO(n)\xi\colon B \to BO(n), the space of BB-structured invertible TFTs valued in a spectrum XX is XhBX^{hB}, where BB acts via ξ\xi followed by ⊖Jn\ominus J_{n}. Proposition 4.3 determines this action on πkS\pi_{k}S for k≤3k \le3. For X=SX = S, this provides the coefficient module for the homotopy fixed point spectral sequence computing ShBS^{hB} in low degrees.

Physical Interpretations

In this section, we briefly indicate how our main theorem (Theorem 3.18) connects to the physics of invertible topological phases and anomalies. The discussion is intentionally informal and aims to translate our mathematical results into the language of field theory; we make no claim to physical rigour. Conjectural statements are explicitly marked.

Invertible field theories and anomalies

An invertible nn-dimensional topological field theory is one where the state spaces are invertible under the tensor product. In physics, invertible TFTs describe the low-energy limits of symmetry-protected topological (SPT) phases and the anomalies of quantum field theories; see [19, 10, 7, 13] for the physics. Mathematically, they are classified by the cobordism hypothesis: for a tangential structure B→BO(n)B \to BO(n), the space of BB-structured invertible TFTs valued in a connective spectrum XX is XhBX^{hB}, where BB acts on XX via ξ\xi followed by ⊖Jn\ominus J_{n} (with our convention). In the lowest dimensions, the classification becomes concrete, and we unpack it below.

Spacetime dimension 1: time-reversal and fermion parity

For n=1n = 1, O(1)≅Z/2O(1) \cong\mathbb{Z}/2, generated by a reflection. By Proposition 4.3, the action on πkS\pi_{k}S is multiplication by −1-1:

π0S≅Z\pi_{0}S \cong\mathbb{Z}: x↦−xx \mapsto-x, fixed subgroup {0}\{0\}.

π1S≅Z/2\pi_{1}S \cong\mathbb{Z}/2: trivial, fixed subgroup Z/2\mathbb{Z}/2.

π2S≅Z/2\pi_{2}S \cong\mathbb{Z}/2: trivial, fixed subgroup Z/2\mathbb{Z}/2.

π3S≅Z/24\pi_{3}S \cong\mathbb{Z}/24: x↦−xx \mapsto-x, fixed subgroup Z/2\mathbb{Z}/2.

Physically, O(1)O(1) is time-reversal symmetry. The invariant on π0\pi_{0} classifies the gravitational response: theories with nonzero π0\pi_{0}-component are not time-reversal symmetric, reflecting the fact that time-reversal complex conjugates the partition function on a closed 1-manifold. The Z/2\mathbb{Z}/2 invariant on π1\pi_1 (the η\eta-invariant) is time-reversal invariant, consistent with the fact that the Kitaev chain and other 1d fermionic SPT phases are protected by time-reversal. The Z/24\mathbb{Z}/24 invariant on π3\pi_3 has a Z/2\mathbb{Z}/2 subgroup fixed by time-reversal, corresponding to the bosonic SPT phases that survive time-reversal symmetry.

Remark 5.1. The fact that the π1≅Z/2\pi_1 \cong\mathbb{Z}/2 invariant is fixed by time-reversal matches the physics of the Kitaev chain: the boundary of the nontrivial phase is a 0 + 1d system whose effective description is captured by this invariant. Our computation takes place in the fully framed setting, which captures the gravitational (framing) anomaly sector. A complete treatment of fermionic SPT phases would replace the framed bordism category with a spin or pin±^{\pm} bordism category; our main theorem applies to any tangential structure B→BO(n)B \to BO(n) via Corollary 2.8, so such refinements are accessible within the framework of this paper.

Spacetime dimension 2: framing anomaly and the spin statistics connection

For n=2n = 2, O(2)O(2) has π0≅Z/2\pi_0 \cong\mathbb{Z}/2. Rotations act trivially on all πkS\pi_kS; reflections act by multiplication by −1-1, identical to the n=1n = 1 case. The action on low homotopy groups is therefore the same as above. The physically relevant invariants for 2-dimensional framed theories are:

  • ♮1 π0S≅Z\pi_0S \cong\mathbb{Z}: a reflection-invariant theory must have trivial gravitational response (the partition function on a closed 2-manifold is invariant under orientation reversal).

  • ♮2 π1S≅Z/2\pi_1S \cong\mathbb{Z}/2: the η\eta-invariant classifies 2d fermionic invertible phases; it is reflection-invariant.

  • ♮3 π3S≅Z/24\pi_3S \cong\mathbb{Z}/24: this is the framing anomaly (or chiral central charge mod 24). The reflection-invariant phases are those with framing anomaly 0 or 12 mod 24.

Remark 5.2. Although rotations act trivially on each πkS\pi_kS as a self-map, the higher homotopy coherence of the O(2)O(2)-action (Remark 4.4) means that a 2π2\pi rotation acts by η∈π1S\eta\in\pi_1S in the homotopy fixed point spectral sequence. This is the mathematical reflection of the spin-statistics theorem: a 2π2\pi rotation can act non-trivially on the space of states, even though it acts trivially on the partition function (See Remark 5.3 below). For n=2n = 2, this η\eta-action contributes to the d2d_2 differential and can obstruct the existence of rotation-equivariant theories.

Remark 5.3. The framing anomaly of a 2d chiral theory is characterised by the failure of the theory to be defined on an arbitrary framed manifold without a choice of 3-dimensional bulk. As Witten explains [29], pp. 191–192, the Hilbert space of a chiral CFT depends on a choice of rigging (framing) at the boundary: under a $2\pi$ rotation of the rigging, a state of conformal weight hh transforms as ∣ψ⟩↦e2πih∣ψ⟩|\psi\rangle\mapsto e^{2\pi i h}|\psi\rangle, while the partition function on a closed surface is independent of the rigging. In the framework of this paper, the anomaly of a 2d framed TFT valued in SS is classified by π3S≅Z/24\pi_3S \cong\mathbb{Z}/24: the theory assigns an element of π3S\pi_3S to a framed 3-manifold via the TFT’s extension to the point, and this element is precisely the framing anomaly (or chiral central charge modulo 24). A 2d theory can be realised as the boundary of a 3d invertible theory if and only if its framing anomaly vanishes, recovering the classic anomaly inflow result [28]. The Z/2\mathbb{Z}/2 fixed subgroup of the O(2)O(2)-action on $\pi_3S ( {xref}Proposition 4.3`) means that reflection-symmetric 2d theories can have framing anomaly 0 or 12 modulo 24, while theories with other values of the framing anomaly break orientation-reversal symmetry.

Remark 5.4. Witten, in [28], Equations 2.24, 2.25, 4.41 shows that under a change of framing by ss units, the partition function transforms by the phase exp⁡(2πis⋅c/24)\exp(2\pi i s \cdot c/24), so the framing anomaly is classified by Z/24\mathbb{Z}/24.

Spacetime dimension 3 and beyond

For n≥3n \ge3, the O(n)O(n)-action on πkS\pi_kS is identical to the n=2n=2 case: rotations act trivially, reflections act by multiplication by −1-1. The physical interpretation of the low-degree invariants extends straightforwardly from the 2-dimension discussion.

The higher homotopy coherence becomes richer in dimension 3: the JJ-homomorphism on π3(SO(n))\pi_3(SO(n)) (for n≥3n \ge3) involves the element ν∈π3S≅Z/24\nu\in\pi_3S \cong\mathbb{Z}/24, which enters the homotopy fixed point spectral sequence through higher differentials [4]. This suggests that 3-dimensional invertible phases with crystalline symmetries may exhibit more subtle mechanisms for anomaly cancellation.

Relation to the Adams Conjecture and Chromatic Phenomena

The JJ-homomorphism is intimately related to the Adams Conjecture and the image of JJ in the stable homotopy groups of spheres [4]. Our main theorem shows that the O(n)O(n)-action on the space of invertible TFTs valued in SS factors through the image of ⊖J\ominus J. This suggests that the classification of SPT phases is controlled by the chromatic filtration of the sphere spectrum: the image of ⊖J\ominus J captures the height-1 chromatic layer, while higher chromatic phenomena (detected by Morava EE-theories) could classify more exotic invertible phases. Amartya Shekhar Dubey, Department of Mathematical Sciences, Norwegian University of Science and Technology, Trondheim, Norway

Email address: [email protected]

Luuk Stehouwer, Durham University, Department of Mathematical Sciences, Upper Mountjoy, Durham DH1 3LE, UK

Email address: [email protected]

References

  1. [1]Matthew Ando, Andrew J. Blumberg, David J. Gepner, Michael J. Hopkins, and Charles Rezk. Units of ring spectra and Thom spectra, 2009.arxiv.org/abs/0810.4535
  2. [2]Matthew Ando, Andrew J. Blumberg, David Gepner, Michael J. Hopkins, and Charles Rezk. An ∞-categorical approach to R-line bundles, R-module Thom spectra, and twisted R-homology. J. Topol., 7(3):869–893, 2014.
  3. [3]Matthew Ando, Andrew J. Blumberg, David Gepner, Michael J. Hopkins, and Charles Rezk. Units of ring spectra, orientations and Thom spectra via rigid infinite loop space theory. J. Topol., 7(4):1077–1117, 2014.DOI
  4. [4]J. Frank Adams. On the groups J(X). In Differential and Combinatorial Topology (A Symposium in Honor of Marston Morse), pages 121–143, 1963.
  5. [5]Benjamin Antieau and David Gepner. Brauer groups and étale cohomology in derived algebraic geometry. Geom. Topol., 18(2):1149–1244, 2014.DOI
  6. [6]M. F. Atiyah. Thom complexes. Proc. London Math. Soc. (3), 11:291–310, 1961. Available at https://webhomes.maths.ed.ac.uk/~v1ranick/papers/atiyahthom.pdf.DOI
  7. [7]Agnès Beaudry and Jonathan A Campbell. A guide for computing stable homotopy groups. arXiv preprint arXiv:1801.07530, 2018.arxiv.org/abs/1801.07530
  8. [9]Damien Calaque and Claudia Scheimbauer. A note on the (∞, n)-category of cobordisms. Algebr. Geom. Topol., 19(2):533–655, 2019.DOI
  9. [10]Arun Debray. Invertible phases for mixed spatial symmetries and the fermionic crystalline equivalence principle. arXiv preprint arXiv:2102.02941, 2021.arxiv.org/abs/2102.02941
  10. [11]Daniel S Freed and Michael J Hopkins. Reflection positivity and invertible topological phases. Geometry & Topology, 25(3):1165–1330, 2021.DOI
  11. [12]Daniel S. Freed. The cobordism hypothesis. Bull. Amer. Math. Soc. (N.S.), 50(1):57–92, 2013.arxiv.org/abs/1210.5100
  12. [13]Daniel S Freed. Anomalies and invertible field theories. In Proc. Symp. Pure Math, volume 88, page 25, 2014.arxiv.org/abs/1404.7224
  13. [14]David Gepner and Rune Haugseng. Enriched ∞-categories via non-symmetric ∞-operads. Advances in Mathematics, 279:575–716, 2015.DOI
  14. [15]Davide Gaiotto and Theo Johnson-Freyd. Condensations in higher categories. arXiv preprint arXiv:1905.09566, 2019.arxiv.org/abs/1905.09566
  15. [16]Daniel Grady and Dmitri Pavlov. The geometric cobordism hypothesis. arXiv preprint arXiv:2111.01095, 2021.arxiv.org/abs/2111.01095
  16. [17]Søren Galatius, Ulrike Tillmann, Ib Madsen, and Michael Weiss. The homotopy type of the cobordism category. Acta Math., 202(2):195–239, 2009.arxiv.org/abs/math/0605249
  17. [18]Antoni A Kosinski. Differential manifolds. Courier Corporation, 2013.
  18. [19]Anton Kapustin and Ryan Thorngren. Anomalous discrete symmetries in three dimensions and group cohomology. Physical review letters, 112(23):231602, 2014.pubmed.ncbi.nlm.nih.gov/24972194
  19. [20]Jacob Lurie. On the classification of topological field theories. In Current developments in mathematics, 2008, pages 129–280. Int. Press, Somerville, MA, 2009.arxiv.org/abs/0905.0465
  20. [21]Jacob Lurie. Higher Algebra. September 2017. https://www.math.ias.edu/~lurie/papers/HA.pdf.
  21. [22]JP May. The Geometry of Iterated Loop Spaces. 2007.
  22. [23]Hoang Kim Nguyen, George Raptis, and Christoph Schrade. Adjoint functor theorems for ∞-categories. Journal of the London Mathematical Society, 101(2):659–681, 2020.DOI
  23. [24]Sergio Pallante. Green anguish. 2026 Joint Mathematics Meetings Art Exhibition, 2026. Online; accessed 14 November 2026.
  24. [25]Douglas C Ravenel. Complex cobordism and stable homotopy groups of spheres. American Mathematical Soc., 2003.DOI
  25. [26]Christopher Schommer-Pries. Invertible topological field theories. J. Topol., 17(2):Paper No. e12335, 64, 2024.
  26. [27]Germán Stefanich. Presentable (∞, n)-categories. arXiv preprint, arXiv:2011.03035, nov 2020. Version 1.arxiv.org/abs/2011.03035
  27. [28]Edward Witten. Quantum field theory and the Jones polynomial. Communications in mathematical physics, 121(3):351–399, 1989.
  28. [29]Edward Witten. On holomorphic factorization of WZW and coset models. Communications in Mathematical Physics, 144(1):189–212, 1992.DOI

Paper details

vlnc:W_6uxefygp
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents